Israel Economic Review Vol. 10, No. 1 (2012), 39–94
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL
AND ESTIMATING PRICE ELASTICITY
DORON SAYAG*
Abstract
This paper studies price movements over the past decade in the Israeli housing
market at the regional level and analyzes the variables that affect housing
prices and the estimation of their elasticity. The price indices were calculated
based on the hedonic methodology, using the CPIM technique (tracking a
representative house). The selected model was adapted to the Israeli market
after conducting numerous simulations to optimize the structure of the
function and the number and type of explanatory variables.
The basic research assumption posits that the housing market does not
consist of a single market, but of several submarkets having mixed price
trends that differ in their intensity. The existence of varying price trends in
different geographical areas stems from differential levels of regional demand
deriving from a range of parameters, such as the local unemployment rate,
regional disposable household income, regional balance of migration and so
forth. On the supply side, regional parameters affecting the housing market
include the unsold inventory level, the number of housing starts, etc.
The measurement was carried out over the years 1999-2009 in nine
geographical submarkets in Israel encompassing 64 urban communities, using
a least-squares multivariable hedonic regression analysis, and it explains, on
average, more than 70% of the variance in prices of houses by means of eight
explanatory variables. The majority of the explanatory variables included in
the study were found to have a level of significance of 5% or less.
Key words: hedonic price, price index, quality adjustment, hedonic regression, utility
function, measurement by the least-squares method.
1. INTRODUCTION
The real estate market in general and the residential real estate market in particular reflect
the wealth and health of the national economy. In a developed market, increases in housing
prices usually reflect expectations for economic growth, a high level of job security and so
*
The Hebrew University and the Central Bureau of Statistics. Email: dorons@cbs.gov.il
40
ISRAEL ECONOMIC REVIEW
forth, whereas price decreases reflect expectations for economic recession and a low level
of job security.
The real estate sector is of such central importance to the national economy that it is
customary to attribute reciprocal influences to the two. Thus, the housing market is
impacted by the state of the economy and the macroeconomic climate (interest rates,
unemployment, inflation, average wage, exchange rates, tax rates, etc.), but at the same
time exerts a considerable influence on the economy.
The real estate sector, which is considered the "engine of the economy," pulls after it
professionals from a wide range of fields, such as entrepreneurs, bankers, architects,
appraisers, surveyors, lawyers, insurers, brokers, electricians, floor layers, carpenters,
glaziers, plumbers, transporters and other professionals from the different fields of
construction. A flourishing real estate sector can be a catalyst for national economic growth
and generate a chain reaction, with an increase in employment in the related sectors as well
as higher government revenues from land taxes which percolate into other sectors.
The official information on housing prices in Israel, published by the Central Bureau of
Statistics, includes a nationwide price index and average quarterly prices by geographical
area. These data do not provide all the knowledge needed to gain familiarity with the Israeli
housing market and for analyzing the influences and implications of public reforms and
macroeconomic changes. Furthermore, the lack of regional information based on empirical
data and on accepted scientific methodologies causes interested parties to publish data that
serve their interests and sometimes mislead the public.
This paper studies movements in housing prices over the past decade and analyzes the
effects of house characteristics on housing prices. The research relies on a broad database
including more than 700,000 observations spread over the years 1999-2009.
In addition, the study for the first time measures empirically the regional variation in
prices of houses. The measurement was carried out in nine geographical submarkets in
Israel encompassing 64 urban communities,
The motivation for a differential regional estimation is the fact that the housing market
consists of submarkets that are differentiated from each other by their geographical
location. This differentiation in the submarkets is evidenced by the price gaps between the
geographical areas arising from the variation in demand for housing between the different
areas and the impossibility of relocating houses. The existence of differential price trends is
examined in this study on an empirical basis by regional estimation.
Besides measuring the movements in house prices, the study also analyzes the factors
affecting house prices and estimates price elasticity taking into account house
characteristics.
The characteristics affecting housing prices are divided in the study into two categories:
i) Environmental characteristics – Variables characterizing the environmental
quality of the house, including the socioeconomic level of the area's residents, the
proximity to employment centers and the level of community services, represented by the
use of the peripherality index- accessibility component, and the degree of the security risk
in communities on the Lebanese border and near the Gaza Strip, which are referred to as
confrontation line communities.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 41
ii) House quality characteristics – Variables characterizing the quality of the house,
including number of rooms, net house size, house type and house age.
The factors affecting housing prices were analyzed using the hedonic regression method
which enables estimating the effect of each unit of a characteristic on the price.
2. DIFFICULTIES AND METHODS OF ESTIMATING HOUSE PRICE MOVEMENTS
a. Difficulty in Measuring House Price Movements
The accepted method for measuring price movements is by building a price index. Price
indices provide a numerical expression for changes in the price of a good or a service over
time, enabling a comparison between price levels in different periods.
The purpose of building a housing price index is to create a continuous time series of
changes in the prices of houses (new and secondhand) over time.
The measurement relied on actual completed transactions between a willing seller and a
willing buyer. The main reason for using actual completed transactions is the fact that there
is no price for a house prior to its actual sale.
The price for a house requested by the seller cannot serve as the information for a
database measuring price movements, since it is merely a theoretical price requested by the
seller. Furthermore, the gap between the price requested by the seller and the price at which
the house is actually sold changes from one period to another and between geographical
areas (according to demand) and also depends on personal characteristics of the seller,
which are irrelevant to the present study.
Difficulty in measuring house prices
The difficulty in measuring house prices stems from two factors:
1. The great heterogeneity in house characteristics.
2. Timing differences between houses of equal quality.
Both the above differences are strongly reflected in the database available to the
researcher.
To illustrate the difficulty in measuring the movement in house prices, a comparison
was made between the structure of the database for a consumer good sold in retail stores
and the structure of the database for housing prices.
Database for consumer goods
The database for consumer goods includes prices from observations in retail stores that are
collected at fixed time intervals. Since the prices are determined by the supply side (the
individual consumers are too insignificant to affect the price), they can be collected from
the retail stores at fixed time intervals.
42
ISRAEL ECONOMIC REVIEW
The observations collected represent a sample of the goods existing in the market and
include information on the good's characteristics, such as the manufacturer's name, the type
of product, the product's weight or content, etc.
The database for such a consumer good appears as a matrix of the collecting periods, the
types of products and their prices.
Table 1
Database for Consumer Goods
Type of
Product
A
t=1
PA1
t=2
PA2
Period
t=3
PA3
t=4
PA4
t=5
PA5
B
PB1
PB2
PB3
PB4
PB5
C
PC1
PC2
PC3
PC4
PC5
D
PD1
PD2
PD3
PD4
PD5
E
PE 1
PE 2
PE 3
PE 4
PE 5
Database of houses
As noted, the database of houses depends on actually completed sales, and therefore there
can be no fixed time intervals for collecting prices of houses of equal quality. Furthermore,
the combinations of house characteristics are so varied that it is virtually impossible to find
apartments of equal quality.
The database of houses is illustrated in the following table:
Table 2
Database of Houses
Types of Houses
t=1
C
Period
t=3
t=4
P
1
B
P
PC3
PD5
D
…∞
t=5
2
A
A
B
t=2
P∞4
The database described above requires special methodologies to enable a price
comparison. In other words, in order to estimate the movement in prices for different
quality houses, one must first reduce the houses sold in each period to an equal standard of
quality. The process of standardizing house quality characteristics to obtain a quality
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 43
constant, enabling a price comparison, is termed in the professional jargon "quality
adjustment."
b. Quality Adjustment Methods in the Housing Market
The purpose of price indices is to reflect changes in the price of a product or group of
products, with quality held constant. Failure to make adjustments to eliminate the effect of
changes in price that arise from change in the quality of the product will cause the index to
be biased.
A consensus has prevailed among economists for dozens of years that changes in
quality are a prime source of error in price indices. One reason for a decline in the quality
of price indices could be their failure to reflect the full extent of such changes.
The problem of bias due to quality changes was addressed by three American
committees: the Stigler Committee (1961), the Boskin Committee (1996) and the
Committee on National Statistics (Schultz Report, 2002), each of which found flaws in the
differentiation between changes resulting from quality versus price changes.
The housing market, which is characterized by great heterogeneity of the characteristics
affecting prices, does not allow for the measurement of price trends based on a comparison
of average prices of houses sold. This is because a difference between periods in the
quality mix of houses could affect the results of the measurement and produce a biased
index. The use of a median price is also not helpful, as there is no guarantee that the quality
of the median house will remain constant over time. Moreover, the use of a median price
does not reflect changes in the prices of houses above and below the median. An increase in
the price of luxury houses or of houses in exclusive or high-demand locations will usually
not be reflected in the median price, which is based on a single price. On the other hand, an
increase in the relative number of houses sold, for example, in the peripheral regions will
bring down the median price, irrespective of the actual price trend.
The professional literature suggests several methods for dealing with changes in quality.
Below are the three most widely accepted methods of adjustment for quality changes in the
housing market.
(i) Equivalent-Quality Clusters (Stratification Method)
This method uses a matrix consisting of clusters of houses of equivalent quality. The
division into clusters is performed on the basis of such criteria as region, house type,
number of rooms, year of construction, etc.
Changes in house prices are measured by multiplying the mean price in each cluster by
its relative weight and comparing the result to the previous period.
R
t
=
∑ P
∑ P
j
t
j
t −1
*W
*W
j
j
44
ISRAEL ECONOMIC REVIEW
Where:
Rt = Average change in prices in period t compared to period t-1
Wj = Weight of cluster j, where
∑
W
j
= 1
Pj = Mean price for cluster j
Limitations of the method:
The main limitation of this method lies in the practical aspect. The heterogeneity in regional
and house characteristics is so great that virtually every house differs from the next. This
would necessitate the use of a vast number of clusters (that is, multiple combinations of
different sets of values of the characteristics), to the extent that it would be impossible to
estimate a mean house price for each cluster.
To overcome this practical difficulty, the number of characteristics included in each
cluster has to be reduced; however, this impairs the homogeneity of each cluster in terms of
quality.
(ii) Repeat Sales Method
The repeat sales method estimates the change in prices using a database consisting of
houses that were sold at least twice.
The model assumes that if Pt is the price of the house in period t, then the price in
period t+1 is Pt (1+αt), where αt is the rate of change between the periods t and t+1.
In period t+2 the rate of change will be (1+αt) (1+αt+1) and so forth.
If a house was sold once in period t at a price of Pt and once in period t+j at a price of
t+j
P , then the rate of change in the price between t and t+j+1 will be:
p t + j +1
=α t+j+1
pt
Graph 1
Measurement of Change in House Prices by the Repeat Sales Method
Price P
Pt+j+1
Pt
t
α t+j+1
t +j+1
Time t
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 45
Since there is generally a large time interval between two points of sale in the house
market, the following algorithmic formula is used:
t+ j
t
p t + j +1
i
log( t ) = ∑ log(1 + α ) = ∑δ i Di
p
i =t
i =1
Where δ=log(1+α) and D is a dummy variable equal to 1 for the periods between t and
t+j+1 and 0 for other periods.
Advantages of the method:
1. Overcoming almost completely the problem of a change in quality (since the price
comparison is performed for the same house in different periods).
2. There is no need to use data regarding the characteristics of the house or data of the
geographical area.
Drawbacks of the method:
1. The assumption that the house's characteristics do not change over time does not take
into account circumstances such as aging/renovation of the house, improvement and
deterioration in the residential area, etc.
2. The assumption that the subsample of resold houses represents the overall change in
prices does not take into account the fact that the duration of ownership of less expensive
houses is shorter, which makes their influence on the result greater than their relative
proportion.
3. The method does not enable the measurement of new houses (since they were not
previously sold).
4. The method presents the change in house prices over time and does not provide
information on price levels.
5. There is a difficulty in the single-value identification of the house.
6. The method necessitates a large quantity of repeat sales of houses in every month.
(iii) Hedonic Price Method
The hedonic price method uses a multivariable regression to estimate the effects of quality
variables on price and enables the isolation of these effects. The hedonic hypothesis states
that "heterogeneous goods are aggregations of characteristics" (Triplett, 1988, p. 630).
The ability to control the effects of house characteristics enables estimating the change
in house prices over time according to the periods in which the houses were sold.
Since the beginning of the 1990s, use of the hedonic method for measuring changes in
house prices has increased, and has become the standard method for handling the problem
of heterogeneity in house characteristics.
46
ISRAEL ECONOMIC REVIEW
Advantages of the method:
1. It enables measuring changes in prices of houses of varying quality in a range of
characteristics and in all combinations.
2. Identification of the characteristics influencing house prices and estimation of the
shadow price (in the additive model) or the elasticities (in the logarithmic model).
3. Estimation of the consumer's willingness to pay for each characteristic (Willingness
to Pay – WTP).
4. Ability to broadly analyze the macroeconomic factors influencing house prices.
Drawbacks of the method:
1. Since the house characteristics and geographical area characteristics, which
determine the quality and price of the house, are sold together with the house, there is a
certain difficulty in isolating the marginal addition to the price due to one or another
characteristic.
2. A hedonic price index could be biased when some of the house characteristics are
not observed.
The problem can stem from a lack of data on the characteristics of a house that affect its
price. This may result in the performance of incomplete quality adjustment. The problem
with the hedonic method was noted by Court (1939), Griliches (1961) and Triplett (1969),
but has remained unsolved.
One possible way of minimizing the problem is by combining the repeat sales method
with the hedonic method.1
3. THE HEDONIC MODEL
a. The Hedonic Model – Literature Review
Hedonic price models are a widespread economic tool for investigating consumer
preference of a product's characteristics estimating the influence of the product's
characteristics on its price. The word "hedonic" derives from the Greek word "Hedonikos,"
which means pleasure. In the economic context, the word "hedonic" refers to the benefit the
individual derives from consuming the product's characteristics.
The method was first applied in the 1920s by Haas (1922), who used hedonic prices in
his research towards his Master's thesis. Haas estimated the prices of agricultural land in
Minnesota (USA) during 1916-1919 according to the characteristics of the land (proximity
to the market, depreciated cost of the buildings on the land, fertility of the land, etc.), using
a multivariable regression.
Waugh (1928) developed the fundamentals of the hedonic methodology. In his study
Waugh estimated by a linear regression the correlation between the prices of vegetables
1
For elaboration see: Shepperd, S. (1999).
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 47
(asparagus) and their characteristics, determining the factors that affect these prices and the
extent of this effect.
Court (1939) was the first economist to coin the term "hedonic price." Court, an
economist at General Motors, sought for a model to compare the prices of cars
manufactured at different times. In his study he posited that car size, engine power, etc.,
change from year to year, and therefore it was necessary to "keep" these variables constant,
in order to determine the "pure" price change. Court was the first to develop the concept of
estimating the effects of car characteristics on price for building a quality-adjusted temporal
price index. Court also was the first to develop the Time Dummy method.
Stone (1956) and Griliches (1961) used the Time Dummy method in a number of areas.
Their work "breathed new life" into the hedonic price theory, after this approach failed to
gain traction during many years. In the wake of their study, the Time Dummy method
became the most prevalent for building indexes during the 1960s and 1970s. Griliches
(1961 and 1971) was the first to use other methods not based on the Time Dummy method.
He discussed the advantages of alternative methods for building hedonic indexes and was
named the father of hedonic modeling in the modern era.
In his paper from 1971, Griliches was the first to describe the Characteristics Price
Index method.
Lancaster (1966) attempted for the first time to formulate a mathematical model for a
consumer who consumes one unit or one composition of a product that links the benefit to
the consumer to the product's price and characteristics, assuming a linear relationship.
Muellbauer (1974) presented a model describing optimal behavior for maximizing the
benefit, representing the consumers' "hidden" preferences. In this paper Muellbauer notes
that the model ignores another side of the market, namely, the manufacturer.
Rosen (1974) elaborated on the theory, stressing the importance of hedonic analysis for
manufacturers as well, in view of their interest in manufacturing products with
characteristics that satisfy consumer desires. Rosen developed a hedonic price model for the
estimation of characteristics as a function of supply and demand.
b. The Hedonic Pricing Model – Theoretical Basis
The hedonic pricing model relates to a product as a "bundle" of characteristics, in which the
price of a product is composed of the value of its constituent characteristics – in terms of
the benefit to the user or in terms of the cost, or in terms of both together.
N
P (z) =
∑
pizi
i=1
Where:
P(z) – The price of the product
Zi
– The characteristics vector of the product
Pi
– The hedonic prices of the characteristics
48
ISRAEL ECONOMIC REVIEW
The hedonic model is based on the economic theory of maximization of the utility for
the consumer deriving from a product's characteristics, which was developed by Lancaster
in an article published in JPE 1966. This article presents for the first time consumer utility
as a direct function of a product's characteristics (Zi), i.e.: U=u(Z1,…,ZN).
The Housing Market
Housing characteristics are considered desirable if they increase the level of consumer
utility, and undesirable if they impair the level of consumer utility. According to the theory,
the price of one house relative to another will differ due to the different number of desirable
or undesirable unit characteristics existing in each house.
Each house can be described by means of quality characteristics. For example:
Area of the house in sq.m.
+ (Garden size in sq.m.
+ (Number of car parking spaces
+ (Elevator (yes/no)
+ Other characteristics2
x
x
x
x
Hedonic price per sq.m. in a particular area
Hedonic price per sq.m. of garden
)
Hedonic price per parking space
)
Hedonic price per elevator
)
House price (= aggregated price of the characteristics)
The derivative (slope) according to any characteristic calculated at the consumer's
optimum point is equal to the "price" the consumer is willing to pay for a change in a unit
of the characteristic.
In order to estimate the shadow prices (the change in the house price after the addition
of a unit characteristic Zi, when the rest of the characteristics remain unchanged) of the
house characteristics, an additive function must be used.
Use of the following basic model will give us the added price for each additional room:
(House Price) = β0 + β1(Rooms) + (others factors)
Estimation of the coefficients in a multiplier function will yield the price elasticity
relative to the house characteristics.
c. Hedonic Indices
The professional literature on hedonic price indices discusses two main techniques:
(i) "Time-Dummy Method" (TDM)
This method is based on an estimation of hedonic price equations in which the time appears
directly as a dummy variable. The method is also referred to as the "direct method."
2
In many cases there is importance to the interaction between the explanatory variables – for example,
when using an interaction between the floor and the "elevator" dummy variable: the hedonic price for a
house without an elevator that is located on an upper floor differs from the shadow price for a house that is
located on a lower floor.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 49
The TDM methodology is based on pooled data, which are cross-sectional data over a
number of periods. Estimation of the effect coefficients of the geographical area
characteristics and the house characteristics is done by regressing the price over several
periods.
The house price is explained as a function of several explanatory variables that describe
the effect of the house quality on the price (the "hedonic" variables), and time variables
(dummy variables) that describe the change in prices over time.
In this method, the coefficients of the quality variables are kept constant in every
regression that is run, and the "pure" change in prices is expressed by means of the time
variable coefficients.
The function model3:
K
ln( P i t ) = β
0
+
∑
T
β
k
Z
i,k
k =1
+
∑
δ
t
D
t
+ ε
i
t=1
Where:
ln( Pi t ) – The natural log of the i house price.
– The k quality characteristic.
βk
Z i, k
δ
t
Dt
– The k quality characteristic of the i house.
– The change in the price log between two periods, when all the quality
coefficients are kept constant,
– Time dummy variable vector
1
If the house price is observed in period t
0
If the house price is observed in another period
Dt =
Estimation of coefficient vectors βk β1, β2, β3,….. enables a price comparison of houses
that differ from each other in the ZK characteristics. We assume that the house
characteristics differ from one house to another, but the effect of each characteristic (β1, β2,
β3,… βk ) remains constant during the estimation period.
K
Since the expression ∑ β k Z i subsumes the changes in house quality in monetary
k
k =1
terms, the change in the house price is reflected in the time variable coefficients
T
∑δ
t
Dt .
t =1
The rationale of the method is clear. If we compare the house prices between period t and
period t-1, for any given quality specification, z, then the ratio is equal to the time
coefficient log exponential ∆P = exp(δt) (Melser 2004).
3
The use of a semi-logarithmic function is solely for the purpose of demonstrating the calculation
methods.
50
ISRAEL ECONOMIC REVIEW
(ii) Characteristics Price Index Method (CPIM)
In this method, the hedonic price index is calculated using a fixed "bundle" of house
characteristics in all the estimation periods. One can use the "bundle" of characteristics
from the previous period (Laspeyres) or from the current period (Paasche). A change in
prices between the two periods, estimated by means of those periods but with different
coefficients (the hedonic prices in each period), reflects the pure price changes.
As noted, according to the hedonic assumption, "heterogeneous goods are aggregations
of characteristics." Hence, in order to compare prices between two periods it is necessary to
use a fixed "bundle" of characteristics from the previous period or from the current period.
The central idea: Comparing the price of the product between two different time
periods when the product's characteristics are "kept" constant (constant quality).
The assumption: The coefficients of the quality characteristics vary over time, that is,
the variance in house prices between two periods stems from changes in the hedonic prices
of the houses.
Definition of the model:
Estimation of the characteristic coefficients for period t and period t+1.
K
ln( P
t
t
0
) = β
+
∑
β
t
k
Z
t
k
β
t+1
k
+ ε
t
i
k = 1
K
ln( P
t+1
) = β
t +1
0
+
∑
Z
t +1
k
+ ε
t+1
i
k =1
Where:
Z kt - The values of the house characteristics in period t.
β kt - The hedonic prices of the house characteristics.
Calculation of hedonic index according to characteristics by the Laspeyres method:
K
∆ P Laspeyres =
0
∑
β i1 z i0
∑
β
i =1
K 0
0
i
z i0
i =1
In other words, in the Laspeyres index the change in house prices from period t to
period t-1 is the ratio between the expected prices in period t (based on the mean
characteristics in period t-1 and the hedonic prices from period t) and the actual prices in
period t-1.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 51
Similarly, one can build an alternative index based on a "representative" house in the
current period that is analogous to the Paasche method4.
Calculation of hedonic index according to characteristics by the Paasche method:
1
K
∆ P Paasche =
∑
β
1
i
z
1
i
∑
β
0
i
z
1
i
i = 1
K 1
i =1
The Laspeyres index does not take into account the fact that when the prices for certain
products/ characteristics rise, the consumer substitutes the products/characteristics with
cheaper products/characteristics.
This problem can be overcome by means of a superlative index5.
K1
∆P Fisher =
∑
K
β i1 z i1
*
i =1
K1
∑β
0
i
z
1
i
i =1
0
∑
β i1 z i0
∑
β
1/ 2
i =1
K 0
0
i
z i0
i =1
The Fisher index6 is based on a geometric average between Laspeyres (the left member)
and Paasche (the right member).
d. Comparing the Methods
The TDM technique has been criticized, beginning already in the 1970s by Griliches. There
are several aspects on which the critics base their preference for the CPIM technique:
1. Noncompliance with price index axioms
There are several axioms that form the basis for any price index. The first is the
Monotonocity in Current Prices Axiom, meaning:
I( P tx, P t-1, Z t, Z t-1) > I( P t, P t-1, Z t, Z t-1) ,
if P tx > P t
In other words, if the price of one of the products in the current period increases while
the rest of the prices remain unchanged, the index must also increase. The TDM does not
necessarily comply with this axiom.7
4
The method is consistent with the way in which a consumer/manufacturer price index is estimated—
determining a basket of products in the base period and estimating the variance in the product prices that
were defined in the base period. The only difference is that instead of determining a "representative" basket
of products, a representative "bundle" of characteristics is estimated.
5
There are three superlative indices: Tornqvist, Fisher and Walsh.
6
Since data exist on the number of transactions performed in the current period, the Paasche index can be
used.
7
For elaboration see: Melser Daniel (2005).
52
ISRAEL ECONOMIC REVIEW
2. Index formula – The index formula in the CPIM technique completely separates
between the structure of the hedonic function and the structure of the index function. This is
desirable both theoretically and practically. The method enables building a superlative
index without forcing a superlative structure on the hedonic function (Diewert, 1976).
3. Theoretical justification – Estimating the hedonic price changes (characteristic
coefficients) between the periods is essential for estimating the price variance using the
CPIM technique. Estimation by the TDM technique paradoxically forces the hedonic prices
to remain constant throughout the periods. This forcing of the hedonic prices to remain
constant has been the object of strong criticism by many economists.
Pakes (2003) estimated by a hedonic regression the variance in mobile computer prices
for the years 1995-1999. He rejected the assumption of constant coefficients between the
periods and recommended not using the TDM technique: "…since hedonic coefficients
vary across periods it [the T.D Method] has no theoretical justification."
4. Gaps in results – Dulberger (1989) measured several hedonic indices for mainframe
computers. The results in a comparison between the TDM technique and the CPIM
technique amounted to approximately 2 percent per annum. Other studies (Okamoto &
Sato, 2001) also showed differences in the results of the two methods. As noted, the gaps
stem from forcing the hedonic prices to be the same in the several periods8.
E. Structure of the Hedonic Function
The structure of the function determines the definition and meaning of the hedonic price.
The accepted functions for hedonic measurement are linear functions, semi-logarithmic
functions and double log functions.
The professional literature does not indicate a preference for a particular structure of the
hedonic function. Nevertheless, there is agreement that in the case of the housing market
the double log function and the semi-logarithmic function are to be preferred to the linear
function. This preference is based on empirical studies but also relies to a considerable
degree on theoretical explanations.
Below are the arguments for preferring the double log or semi-logarithmic function
structure in the housing market:
1. Hedonic estimation using a linear function estimates a constant shadow price. In
other words, the coefficients describe the addition to the price (in NIS) for an additional
unit characteristic irrespective of any other characteristics in the house.
The double log and semi-logarithmic model measures the elasticity of each
characteristic and is affected by the general house value.
To illustrate the hedonic price result, differentiation was performed according to the
characteristics of the function in each of the accepted forms. One can see that only in the
double log differentiation does the general price remain within the hedonic price.
8
For elaboration see: Okamoto & Sato (2001); Triplett (2001).
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 53
Table 3
Structure of the Hedonic Functions
Type of Function
Function Structure
Hedonic Price
∂P
= β
∂Z K
K
Linear function
P = β
+
0
∑
βkZ
k
k =1
Semi-logarithmic
function
∑
β k ln Z k
k =1
K
P = β
Double log function
(exponential)
βK
∂P
=
Zk
∂Z K
K
P = β0 +
0
∏
Z
β
K
k
K
k =1
After logarithmic transformation:
K
ln P = ln β 0 +
∑
β
K
ln Z
K
β
∂P
=
∂Z k
Z
k
P
k
k =1
There is no importance to using a hedonic price that includes the product's general
value, for example, in a computer, in which the price of the burner or the screen is not
affected by the computer's other characteristics or by its overall price. By contrast, decisive
importance attaches to this factor in the housing market.
For example: In the linear model, the shadow price (in NIS) for an additional square
meter will be fixed for a new house in an upscale neighborhood and for an old house in a
poor neighborhood. In reality this is unreasonable, because the additional price for a square
meter in a new house in an upscale neighborhood will be higher than in an old house in a
poor neighborhood. The logarithmic model provides a solution to the problem, since it uses
a percentage of the house value and therefore will reflect a higher price for an additional
square meter in a new house located in an upscale neighborhood.
2. Another argument posits that in the majority of explanatory variables in the housing
market model, the willingness to pay for an additional unit characteristic is not fixed but
rather decreases. Therefore, the relationship of these variables is logarithmic. For instance,
the willingness to pay for an additional square meter in a 40 square meter apartment is
greater than in a 150 square meter apartment. In variables in which the marginal willingness
does not decrease, a linear relationship can be used – for example, the variable of
socioeconomic level or degree of peripherality.
4. STUDY DESCRIPTION
a. Description of the Database
The main database used in the study consists of the real estate database system (the
CARMAN system) maintained by the Israel Tax Authority. Information on the purchase of
houses is collected on a regular basis by the betterment tax offices throughout Israel by
means of the CARMAN system. The national file is produced by the Automated Processing
54
ISRAEL ECONOMIC REVIEW
Service (SHAAM) – a computer unit of the Israel Tax Authority that provides computing
services to government tax departments and to external customers – and contains
information on new and secondhand house purchases. The CARMAN data are intended for
statistical reporting and therefore include extensive information on every house that is sold.
This information is based on the details provided in the appreciation tax form which is
submitted by the lawyer in every real estate transaction within 50 days from when the
transaction is completed. The data in the form are entered manually into the CARMAN
system.
b. Study Population
Geographically, the CARMAN data cover all parts of Israel. The study population includes
all houses purchased by private individuals in urban communities only, since the hedonic
pricing method assumes the existence of a competitive market, i.e. multiple buyers and
multiple sellers.
Such a situation exists only in large urban communities, whereas in small communities
or in moshavim (cooperative communities) house supply and demand is limited. The
communities included in the sample are those with 5,000 houses and up.
Table 4
Sample of Communities and Regional Division into Submarkets9
Area
List of Communities Included in the Submarkets
Jerusalem
Tel Aviv
Haifa
Dan
Metropolis
Center
1. Jerusalem
2. Tel Aviv
3. Haifa
4. Bnei Brak
5. Bat Yam
6. Givatayim
9. Or Yehuda
14. Yahud
10. Beit Shemesh
15. Lod
13. Yavne
18. Nes Ziona
19. Petach Tikva
24. Ramle
25. Ofakim
39. Dimona
20. Kiryat Ono
11. Givat Shmuel 12. Taibeh
16. Mevasseret
17. Modi'in
Zion
21. Rosh Ha'ayin 22. Rishon Lezion
26. Eilat
31. Arad
27. Ashdod
32. Kiryat
Malachi
28. Ashkelon
33. Kiryat Gat
29. Beer Sheba
34. Netivot
35. Sderot
36. Hod Hasharon
37. Herzliya
38. Hadera
39. Kfar Saba
40. Netanya
41. Ramat Hasharon
43. Um el Fahm
42, Raanana
44. Tiberias
46. Tamra
48. Karmiel
53. Upper Nazareth
49. Migdal Ha'emek
54. Nesher
45. Tirat
Hacarmel
50. Maalot
55. Acre
47. Yokne'am
Illit
52. Nazareth
57. Pardes
Hanna
58. Safed
61. Kiryat Ata
59. Kiryat Shmona
62. Kiryat Bialik
60. Shfaram
63. Kiryat Yam
South
Sharon
Plain
North
Krayot
7. Holon
51. Nahariya
56. Afula
8. Ramat Gan
23. Rehovot
64. Kiryat Motzkin
9
The number of houses in a city is based on the number of municipal property tax billings. Source of
data: Central Bureau of Statistics, 2003.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 55
c. Potential Explanatory Variables
The potential number of variables affecting the house price is large and varies from one
geographical area to another. Many of these characteristics are not included in the database,
or else the information about them in the database is partial/blank/of low quality.
The present study examined the effect of nine explanatory variables with a theoretical
potential effect on the house price.
Table 5
Description of Potential Variables
Name of Variable
Environmental quality variables
1 Peripherality index – accessibility
component
Description
PAC
2 Socioeconomic level
SEL
3 Security risk – communities on the
confrontation line
CLC
House quality variables
4 House size
5 Number of rooms
6 House age
SIZ
RMS
AGE
7 Special houses
SAP
8 Status: new house or secondhand house
STA
9 Floor
FLR
Component in the peripherality index. Indicates
accessibility to employment areas and
community services.
Socioeconomic level of the population in a
statistical zone.
Israel's northern communities situated within a
radius of up to 9 kilometers from the Lebanese
border, and Gaza-belt communities situated
within a radius of up to 7 kilometers from the
Gaza border.
Net house size
Number of rooms in the house
Difference in years between the transaction
date and the construction date
Dummy variable separating ordinary apartments
in a building and special houses: semidetached
houses, penthouses, villas, garden apartments, etc.
Dummy variable separating secondhand houses
and new houses
Variable indicated the apartment floor
(i) Environmental Quality Characteristics
The housing market differs from other markets in that house prices and price trends may
differ from one geographical area to another. In contrast to other goods that can be moved
from one area to another, the housing market is static, hence the need for using
characteristics that define environmental quality.
The variables examined in the context of environmental quality characteristics (for
some of the geographical areas) are:
• Peripherality index – accessibility component (PAC)
Important influences on house prices stem from the economies of scale of the locality and
its geographical location. A large locality or a locality situated close to a large population
center will benefit from employment areas and diverse and nearby community services,
56
ISRAEL ECONOMIC REVIEW
such as schools, kindergartens and daycare centers, community centers, commercial
centers, culture centers, care centers for the elderly, game courts and sports facilities, health
services, synagogues, etc., and therefore one can expect the prices of houses in large
communities or in communities situated close to population centers to be higher than those
of identical houses in small communities or in communities located at a distance from
population centers.
In order to measure the degree of accessibility to employment areas and community
services, use was made of the potential accessibility component of the peripherality index
published by the Central Bureau of Statistics. The potential accessibility index weights
between the local authority's proximity to all local authorities in Israel and the size of their
population.
252
Pj
The potential accessibility formula:
Ai =
∑
j =1
Where:
Ai – Potential accessibility of local authority
d ij1 . 19
i ;
Pj – Population of local authority j ;
d ij – Distance in kilometers from center of local authority i to center of local authority j ;
d ii – Self-distance of local authority i which is defined as 3 kilometers for any local
authority
The index is divided into 10 clusters, with cluster 1 indicating a low level of
accessibility and cluster 10 indicating a high level of accessibility.
Table 6
Peripherality Index – Potential Accessibility Index10
Geographical Area
Jerusalem
Tel Aviv
Haifa
Dan Metropolis
Center
South
Sharon Plain
North
Krayot
10
Potential Accessibility Level
10 – Jerusalem
10 – Tel Aviv
7 – Haifa
10 – Bnei Brak
8 – Or Yehuda
8 – Yahud
9 – Petach Tikva
7 – Ramle
1 – Ofakim
1 – Dimona
2 – Sderot
8 – Hod Hasharon
8 – Ramat Hasharon
3 – Um el Fahm
3 – Karmiel
4 – Upper Nazareth
1 – Safed
5 – Kiryat Ata
9 – Bat Yam
5 – Beit Shemesh
7 – Lod
9 – Kiryat Ono
10 – Givatayim
9 – Holon
10 – Givat Shmuel 5 – Taibeh
5 – Mevasseret Zion5 – Modi'in
7 – Rosh Ha'ayin 9 – Rishon Lezion
10 – Ramat Gan
6 – Yavne
8 – Nes Ziona
7 – Rehovot
1 – Eilat
1 – Arad
6 – Ashdod
4 – Ashkelon
4 – Kiryat Malachi 3 – Kiryat Gat
4 – Beer Sheba
2 – Netivot
8 – Herzliya
5 – Hadera
8 – Raanana
2 – Tiberias
3 – Tirat Hacarmel
4 – Migdal Ha'emek2 – Maalot
5 – Nesher
4 – Acre
1 – Kiryat Shmona 4 – Shfaram
5 – Kiryat Bialik 5 – Kiryat Yam
Source of data: Central Bureau of Statistics.
8 – Kfar Saba
6 – Netanya
3 – Tamra
3 – Nahariya
3 – Afula
3 – Yokne'am Illit
5 – Nazareth
4 – Pardes Hanna
5 – Kiryat Motzkin
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 57
• Security risk – confrontation line communities (CLC)
Communities situated near the confrontation line are at a greater security risk, a factor
affecting house prices.
Out of the study's population of communities, Kiryat Shmona, Maalot and Nahariya
were defined as northern confrontation line communities, while Sderot was defined as a
southern confrontation line community.
Table 7
List of CLC Communities in the Study11
Locality
1.
Sderot
Distance from Gaza Strip
(km)
3.1
Locality
1.
2.
3.
Kiryat Shmona
Maalot
Nahariya
Distance from Lebanese
Border (km)
3.4
8.5
8.8
• Socioeconomic Level (SEL)
An important characteristic of the geographical area of the house is the socioeconomic level
of the population. The basic units are statistical zones within urban communities with
10,000 residents and up, and smaller communities, which are not subdivided. The statistical
zone was chosen as a geographical basis since it is small enough to be homogeneous but
large enough to enable reliable estimates of socioeconomic characteristics.12 The
socioeconomic level is a value between 1 and 20, with 20 indicating a high socioeconomic
level and 1 indicating a low socioeconomic level.
Setting the socioeconomic level of the geographical unit's residents takes into account
the following variables: demography; standard of living; schooling and education;
employment and unemployment; pension/allowances.
(ii) House Quality Characteristics
The variables included in the house quality characteristics are:
• House size – SIZ: The relevant house size for house buyers is the net area. In cases
where the gross area was indicated without the net area, the net area was imputed based on
the gross area and the type of property.
• Number of rooms – RMS: The number of rooms in a house is defined as the
number of bedrooms + living room. The standard for room size is 8 square meters; a
smaller space is defined as a half room.
11
The distance to the border is calculated from the city center. Source of data: Central Bureau of
Statistics, GIS Unit.
12
One of the problems arising from the use of the socioeconomic index is the difficulty in linking records
from different databases. The link between the socioeconomic index and the houses reported in the
purchase tax file was established by "anchoring" each house to a statistical zone based on the address (city,
street and house number). In cases where no address was found, the house's block-parcel number was used.
58
ISRAEL ECONOMIC REVIEW
• Special apartments – SAP: A dummy variable given a value of 0 for an apartment
in a condominium and a value of 1 for special houses (semidetached, villa, duplex,
penthouse, etc.).
• House age – AGE: The house age is the difference between the year of construction
and the transaction date (rounded to years).
• Status – STA: A dummy variable given a value of 0 for a secondhand house and a
value of 1 for a new house. New house – a house sold up to two years after the construction
date.
• Floor – FLR: A variable indicating the floor on which a purchased apartment is
located.
d. Simulations and Selection of the Optimal Model
Choosing the optimal model involved 32 simulations which examined the level of
significance of each potential variable and its appropriate form (linear of logarithmic). The
simulations were run for the years 1999-2009, on a monthly basis, separately for each
geographical area. The model chosen is a differential model specially adjusted for each
geographical area according to the relevant variables for that area.
The need for a differential model stems from the fact that certain variables are relevant
in some geographical areas but irrelevant in others. For example, the use of the
peripherality index is relevant for the northern, southern, central and other regions, but
irrelevant for areas consisting of a single community or areas in which the peripherality
level is the same for all the communities included in the area, such as the Krayot. Various
area characteristics, such as the floor number, are relevant only for Tel Aviv, which is
characterized by high-rise buildings. Proximity to the Lebanese border or to the Gaza Strip
is a variable relevant solely to the north and the south.
e. Definition of the Selected Model
• Hedonic estimation method – An empirical analysis of the two hedonic models
showed negligible differences between the two. Based on the recommendations of the
professional literature, the CPIM model was chosen.
• Calculation method – Estimation of the coefficients was done by the ordinary least
squares (OLS) method.
Function structure: Semi-logarithmic
• Listing and description of the variables in the model – 8 explanatory variables: 3
environmental quality variables and 5 house quality variables.
The type and number of variables vary between the geographical areas, according to the
following table.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 59
House price in NIS (logarithmic)
Environmental quality characteristics
Socioeconomic level (linear)
Peripherality index – accessibility (linear)
Security risk – confrontation line
communities (dummy)
House quality characteristics
House size (logarithmic)
Number of rooms (logarithmic)
House age (logarithmic)
Special houses (dummy)
Floor (linear)
Krayot
North
Sharon Pl
South
Center
Dan Metro
Haifa
Geographical Areas in Which the
Variables Are Included
Tel Aviv
Jerusalem
Table 8
Description of Explanatory Variables in the Model According to Geographical Area
Variable
Symbol
+
+
+
+
+
+
+
+
+
P
+
+
+
+
+
+
+
+
+
+
+
+
+
+
SEL
PAC
CLC
+
+
+
+
SIZ
RMS
AGE
SAP
FLR
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
General area function equation
Ln
=
Pi
β 6 Ln
Ln β 0 + β 1 SEL i + β 2 PAC i + β 3 CLC i + β 4 Ln
SIZ i + β 5 Ln
RMS i +
AGE i + β 7 SAP i + β 8 FLR i
The price calculated for a house with average characteristics for area j is:
0
( β t + β t SEL j + β t PAC
0
pˆ t
j
= e
1
β t Ln
6
AGE
0
0
t
t
j + β 3 CLC j + β 4 Ln
0
0
SAP j + β t
FLR j )
8
2
0
t
+
β
j
7
SIZ
0
t
j + β 5 Ln
RMS
0
j+
Area index formula:
0
0
0
( β t − 1 + β t − 1 SEL j + β t − 1 PAC j + β t − 1 CLC j + β t − 1 Ln
I t = I tj−1
j
0
1
2
3
0
0
0
β t − 1 Ln AGE j + β t − 1 SAP j + β t − 1 FLR j )
7
6
8
e
0
0
4
0
( β t − 1 + β t − 1 SEL j + β t − 1 PAC j + β t − 1 CLC j + β t − 1 Ln
0
1
2
3
4
β t − 1 Ln
e 6
AGE
0
SIZ j + β t − 1 Ln
5
0
SIZ j + β t − 1 Ln
5
0
RMS j +
RMS
0
j+
0
t − 1 SAP 0 + β t − 1 FLR 0 )
j +β7
j
j
8
Definition of representative house by geographical area
The CPIM technique, which is based on a representative house over time, ensures that the
quality is kept unchanged over time. The calculation of the representative house for each
area was determined based on the mean characteristics of houses sold in 1999. As can be
60
ISRAEL ECONOMIC REVIEW
seen in the following table, the calculation of the representative house for each area
included the nine characteristics that were included in the simulations.
Table 9
Description of Representative House by Geographical Area
Jerusalem
Tel Aviv
Haifa
Dan Metropolis
Center
South
Sharon Plain
North
Krayot
SIZ
RMS
AGE
SEL
PAC
FLR
SAP
CLC
75.9
71.8
73.6
68.4
87.2
85.6
90.4
88.2
78.4
3.3
3.0
3.2
3.1
3.7
3.6
3.9
3.6
3.4
24.2
30.3
27.0
27.3
12.2
8.5
12.4
12.7
20.6
12.5
14.4
13.8
12.7
13.2
8.2
13.2
10.2
10.8
9.5
7.9
4.9
6.9
3.3
-
2.9
-
0.002
0.045
0.028
0.021
0.060
0.148
0.286
0.120
0.090
0.02
0.21
-
5. RESULTS
a. Description of Explained Variance
As expected, the explained variance is not uniform for all the geographical areas. Graph 2
shows the results of the explained variance (R2) in each area. The graph for each area
includes all the monthly results in the years 1999-2009.
Graph 2
Description of Explained Variance by Geographical Areas
It is apparent from the graph that the explained variance in Jerusalem is significantly
lower than in the other areas. It is also apparent that the distance between the minimum and
maximum variance is greater in Tel Aviv.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 61
An examination of the explained variance along the time axis does not show any trend
of improvement or worsening of the explained variance.
It is interesting to note that developments in Tel Aviv and Jerusalem are similar, while
in Haifa the picture looks different. Perelman (2011)13 shows that the ratio of apartments
stock to households in Haifa is higher when compared to other cities. Note also that after
the economic crisis the decline in Tel Aviv is remarkable when compared to other cities. A
possible explanation (not explored in the paper) is the fall in investors purchases, which is
consistent with the higher stock of apartments for rent in Tel Aviv as compared with
Jerusalem and Haifa: 47.3 percent against 26 and 30.6, respectively (source: expenditure
survey of the CBS).
b. Regional change in house prices
Graph 3
Behavior of Jerusalem Index
80
Pct change (right axis)
Index (left axis)
(‫אחוז שינוי )ציר ימין‬
(‫מדד )ציר שמאל‬
180
70
160
60
Index
Index
140
50
120
40
100
30
80
20
60
40
10
20
0
-10
.
0
Percent Change
200
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Graph 4
Behavior of Tel Aviv Index
80
Pct change (right axis)
Index (left axis)
(‫אחוז שינוי )ציר ימין‬
(‫מדד )ציר שמאל‬
180
160
70
60
50
120
40
100
30
80
20
60
40
10
20
0
-10
.
0
1999
13
2000
2001
Bank of Israel's site, 2011.
2002
2003
2004
2005
2006
2007
2008
2009
‫אחוזשינוי‬
Index
140
Percent Change
200
62
ISRAEL ECONOMIC REVIEW
Graph 5
Behavior of Haifa Index
160
70
60
140
50
120
‫אחוזשינוי‬
Index
80
Pct change (right axis)
Index (left axis)
(‫אחוז שינוי )ציר ימין‬
(‫מדד )ציר שמאל‬
180
40
100
30
80
20
60
40
10
20
0
0
Percent Change
200
.
-10
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Graph 6
Behavior of Dan Metropolis Index
200
80
(‫ימין‬change
‫שינוי )ציר‬
‫אחוז‬
Pct
(right
axis)
(‫מדד )ציר שמאל‬
Index
(left axis)
160
70
60
Index
140
50
120
40
100
30
80
20
60
40
10
20
0
-10
.
0
Percent Change
180
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Graph 7
Behavior of Central Region Index
200
80
Pct change (right axis)
Index (left axis)
(‫אחוז שינוי )ציר ימין‬
(‫מדד )ציר שמאל‬
180
160
70
30
80
20
60
40
10
20
0
0
-10
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
‫אחוזשינוי‬
40
100
.
Index
50
120
Percent Change
60
140
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 63
Graph 8
Behavior of Southern Region Index
Table 10: Annual Price Change 200
80
(‫ימין‬
‫שינוי )ציר‬
Pct
change
(right‫אחוז‬
axis)
(‫מדד )ציר שמאל‬
Index
(left axis)
160
70
60
140
Index
40
100
30
80
‫אחוזשינוי‬
50
120
20
60
40
10
20
0
-10
.
0
Percent Change
180
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Graph 9
Behavior of Sharon Plain Index
200
80
Pct change (right axis)
Index (left axis)
(‫אחוז שינוי )ציר ימין‬
(‫מדד )ציר שמאל‬
180
160
70
40
100
30
80
20
60
40
10
20
0
-10
.
0
‫אחוזשינוי‬
‫מדד‬
Index
50
120
Percent Change
60
140
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Graph 10
Behavior of Northern Region Index
200
80
(‫ימין‬
‫שינוי )ציר‬
‫אחוז‬
Pct
change
(right
axis)
(‫מדד )ציר שמאל‬
Index
(left axis)
180
160
70
50
40
100
30
80
20
60
40
10
20
0
0
-10
.
‫מדד‬
Index
120
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Percent Change
60
140
64
ISRAEL ECONOMIC REVIEW
Graph 11
Behavior of Krayot Index
200
80
(‫ימין‬
‫שינוי )ציר‬
Pct
change
(right‫אחוז‬
axis)
(‫מדד )ציר שמאל‬
Index
(left axis)
180
160
70
50
Index
120
40
100
30
80
20
60
40
10
20
0
-10
.
0
Percent Change
60
140
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Table 10
Annual Price Change – Jerusalem
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. Change
Annual rate of change
Annual Pct. Change
(Nominal)
0.8
-9.2
-1.6
12.8
-4.8
9.5
11.7
2.0
8.3
10.3
13.4
63.3
4.6
Annual Pct Change
(Real)
-1.0
-9.2
-3.0
5.9
-3.0
8.2
9.1
2.1
4.7
6.2
9.1
25.2
2.1
Table 11
Annual Price Change – Tel Aviv
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
0.7
22.5
-20.1
4.8
-5.2
4.7
12.6
5.1
21.6
11.8
5.7
74.3
5.2
Annual Pct Change
(Real)
-1.1
22.5
-21.2
-1.6
-3.4
3.4
10.0
5.2
17.6
7.7
1.7
33.6
2.7
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 65
Table 12
Annual Price Change – Haifa
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
8.0
-9.7
-3.5
3.0
-4.7
-3.5
3.3
-6.5
-7.3
5.3
20.7
1.3
0.1
Annual Pct Change
(Real)
6.1
-9.7
-4.8
-3.3
-2.9
-4.6
0.8
-6.4
-10.4
1.4
16.2
-22.3
-2.3
Table 13
Annual Price Change – Dan Metropolis
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
4.0
-7.7
0.4
7.2
-6.0
0.0
4.1
-4.1
11.0
22.3
18.0
55.4
4.1
Annual Pct Change
(Real)
2.1
-7.7
-1.0
0.6
-4.2
-1.2
1.7
-4.0
7.4
17.8
13.6
19.2
1.6
Table 14
Annual Price Change – Central Region
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
1.5
-6.1
0.2
9.6
-4.3
-1.7
7.9
-5.0
7.4
20.7
19.0
55.9
4.1
Annual Pct Change
(Real)
-0.3
-6.1
-1.2
2.9
-2.5
-2.9
5.4
-4.9
3.9
16.3
14.5
19.5
1.6
66
ISRAEL ECONOMIC REVIEW
Table 15
Annual Price Change – Southern Region
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
-2.6
-5.3
-1.1
3.7
-2.2
1.2
7.1
2.7
2.5
17.0
12.8
39.2
3.1
Annual Pct Change
(Real)
-4.3
-5.3
-2.5
-2.6
-0.3
0.0
4.6
2.8
-0.8
12.7
8.6
6.7
0.6
Table 16
Annual Price Change – Sharon Plain
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
3.7
-6.2
0.3
7.9
-2.7
-0.6
9.5
-4.9
4.3
19.1
17.9
55.6
4.1
Annual Pct Change
(Real)
1.9
-6.2
-1.1
1.3
-0.8
-1.8
7.0
-4.8
0.9
14.9
13.4
19.3
1.6
Table 17
Annual Price Change – Northern Region
Years
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct. Change
(Nominal)
-3.1
-1.4
-5.3
-0.9
-4.9
-3.9
-1.5
-6.7
-2.9
6.6
16.0
-9.6
-0.9
Annual Pct Change
(Real)
-4.8
-1.4
-6.7
-6.9
-1.3
-5.0
-3.8
-6.6
-6.1
2.7
11.7
-30.7
-3.3
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 67
Table 18
Annual Price Change - Krayot
Years
Annual Pct. Change
(Nominal)
-4.1
-1.3
-6.1
-1.2
-3.3
-4.2
-1.0
-6.3
-2.7
6.1
16.2
-9.5
-0.9
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Cum. pct. change
Annual rate of change
Annual Pct Change
(Real)
-5.8
-1.3
-7.4
-7.3
-1.4
-5.4
-3.3
-6.2
-5.9
2.2
11.8
-30.6
-3.3
c. Analysis of Factors Influencing House Prices
The characteristic coefficients (βs) signify the elasticity of the price relative to the
characteristics. Elasticity provides information about the extent of the effect that a change
in a particular characteristic will have on the house price. Elasticity is not affected by the
measurement units, and therefore it is suitable for testing the sensitivity of the house price
to changes in the house characteristics.
It should be noted that elasticity of the characteristics assumes that the rest of the
characteristics remain unchanged. However, changes in the house size and changes in the
number of rooms usually occur together.
Table 19
Elasticity of House Characteristics and Environmental Characteristics by
Geographical Area
ESIZ,P
ERMS,P
EAGE,P
ESEL,P
EPAC,P
ECLC,P
ESAP,P
EFLR,P
Jerusalem
0.7521
0.1472
-0.0024
Tel Aviv
0.7435
0.1744
-0.0272
0.0192
-
-
0.1744
-
0.0346
-
-
0.2323
Haifa
0.0131
0.7520
0.2757
-0.0356
0.0512
-
-
0.2443
-
Dan Metro
0.4544
Center
0.5830
0.3985
-0.0306
0.0307
0.0603
-
0.2539
-
0.3205
-0.0278
0.0363
0.0618
-
0.1982
South
-
0.6368
0.2734
-0.0356
0.0281
0.0781
-0.0649
0.1973
-
Sharon Pl
0.5743
0.2752
-0.0290
0.0263
0.0675
-
0.1619
-
North
0.5992
0.3486
-0.0335
0.0345
0.0821
0.1904
0.2045
Krayot
0.6914
0.2710
-0.0364
0.0342
-
-
0.1462
-
Nationwide
0.6430
0.2761
-0.0287
0.0328
0.0700
-
0.2014
-
Table 19 shows that the elasticity of the characteristics relative to the price is rigid for
all the characteristics (0<|E|<1)
68
ISRAEL ECONOMIC REVIEW
• Effect of house size relative to house price
As expected, the house size affects house prices positively at a level of significance of 1%
in all the geographical areas.
The nationwide mean elasticity of this characteristic on the house price is 0.643. The
significance of this estimate is that an increase of one percent in the house size (the rest of
the house characteristics remain unchanged) will raise the house price by 0.64%. In major
cities, i.e., Jerusalem, Tel Aviv and Haifa, the addition to the price (as a percent) for an
additional square meter is higher than in the other areas.
Note that in the Dan Area the elasticity is lower when compared to other areas, or even
to cities that belong to it, like Tel Aviv. However, we show in the next table that the
elasticity for rooms in this area is higher. A possible interpretation of this result is that in
this area there is a higher sensitivity for rooms.
• Elasticity of house price relative to number of rooms in the house
Also the number of rooms affects house prices positively. The mean elasticity of this
characteristic is estimated at 0.2761.
The significance: A transition from two rooms to four rooms, when the house size and
the other house characteristics remain unchanged, will raise the house price by an average
of 27.6 percent.
• Elasticity of house price relative to house age
House age is the only characteristic in the study with a negative effect on the price. The
nationwide elasticity of the price relative to age is -0.0287. The effect of the house age on
the house price was recorded as negative in all the geographical areas. Jerusalem is
differentiated from all other cities and areas in that the effect of the house age is
significantly lower than in other areas. The low devaluation of old houses is apparently due
to the historical value and unique character of Jerusalem's old houses.
• Elasticity of house price relative to socioeconomic level
The nationwide elasticity of the price relative to the socioeconomic level is 0.0328. The
effect of the socioeconomic level was found to be positive in all the geographical areas,
signifying that an increase by one unit in the socioeconomic level will result in an average
increase of 3.28 percent in house prices.
• Elasticity of house price relative to degree of peripherality of the community
The study's findings show that the nationwide elasticity of the price relative to the degree of
peripherality is 0.07. The effect of the degree of peripherality was found to be positive in all
the geographical areas, signifying that an increase by one unit in the degree of peripherality
will result in an average increase of 7 percent in house prices.
• Elasticity of house price relative to house type
The study's findings show that the nationwide elasticity of the price relative to the house
type is 0.2014.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 69
The significance: A transition from a house in a multi-story apartment building to an
upscale house (detached house, penthouse, villa, etc.) raises the house price by 20%.
• Effect of floor on the house price
The effect of the floor on the house price was examined only in Tel Aviv. The study's
findings show a positive correlation between the floor and the house price, equivalent to an
average increase of 1.315 percent in the house value.
• Effect of proximity to the Lebanese border and to the Gaza Strip on the house
price
Houses located on the northern and southern confrontation lines are subject to two
opposite effects:
(1) The security effect, which operates to lower house prices in these communities due
to a higher level of risk.
(2) The economic effect, reflected in increased demand and higher prices due to tax
concessions of up to 25%.
The study's findings show that in the northern confrontation line communities, the
economic effect is stronger than the security effect, causing house prices to be an average
of 19% above their economic value relative to the environmental characteristics, i.e., the
degree of peripherality and the socioeconomic level.
In the southern confrontation line communities, represented by the city of Sderot, the
security effect was found to be stronger than the economic effect. Thus, the value of houses
in Sderot is lower by 6.5% than their economic value, in spite of the tax concessions.
6. SUMMARY AND CONCLUSIONS
This study joins a long series of studies around the world that support the use of hedonic
prices for measuring movements in house prices and as a tool for studying the effects of
house characteristics on the house price.
Apart from information on changes in house prices, the CPIM technique provides a new
dimension of knowledge on house prices, namely, the monthly price level of a
representative apartment.
The results of the regional house price indices provide empirical evidence of mixed
trends in house prices, which differ from one geographical area to another. These empirical
findings substantiate the study's central thesis, that the residential housing market consists
of several regional submarkets that are characterized by different levels and price trends.
Findings regarding changes in regional house prices:
• During 2008 and 2009, house prices in Israel rose sharply in all the geographical
areas, to rates of between 15% and 40%.
• Price rises in all regions presents a new phenomenon compared to the mixed regional
price trends that characterized Israel's real estate market in all the other periods of the study.
70
ISRAEL ECONOMIC REVIEW
• The areas leading the increase in prices during this period are the Dan Metropolis,
the Central Region and the Sharon Plain.
• In Haifa, the north of the country and the Krayot (Haifa bayside suburbs), the price
rises in the past two years are an adjustment to the extended price decreases during the nine
years preceding the rise. Therefore, the fear of a real estate bubble does not apply to these
areas.
• Prices behaved similarly in the Dan Metropolis, the Central Region and the Sharon
Region. During the period of the study, price increases of more than 50% were recorded,
about three-quarters of the increase occurring during 2008-2009.
• Tel Aviv and Jerusalem recorded the sharpest increases during the period of the
study. In Tel Aviv house prices rose at a rate of 75%, and in Jerusalem at a rate of 65%.
The likelihood of a real estate bubble in these cities is much greater than in other areas of
the country.
• Excluding Tel Aviv, the house market in most areas of Israel does not react
immediately to specific occurrences or economic events. In Tel Aviv, on the other hand,
events such as the crisis in the high-tech industry at the end of the year 2000 and the global
economic crisis at the end of 2008 were reflected in house prices, which reacted
immediately and sharply. It may be presumed that the high percentage of investors in the
city (based on the percentage of rented houses) and the percentage of luxury apartments in
the city render Tel Aviv highly sensitive to economic events.
• The price gaps among the representative houses in the different geographical areas
widened sharply during the period of the study, due to price increases in areas characterized
by high price levels (Haifa, Tel Aviv and the Sharon), versus price decreases in areas
characterized by low price levels (the north and the Krayot). Thus for example, the ratio of
prices between representative houses in Tel Aviv and in the north, which was less than two
times in 1999, rose to more than three times in 2009.
• In the northern and southern regions, which were characterized by similar price
levels in 1999, price levels gradually but systematically began to part ways, with the gap in
representative house prices standing at the end of 2009 at 30%.
• Despite the proximity of Tel Aviv to the Dan Metropolis which envelopes it, there
are behavioral differences in prices. The sharp rise in house prices in Tel Aviv indicates
that from the standpoint of house buyers (owners and investors), the Dan Metropolis is not
a substitute for the city of Tel Aviv.
• A similarity in the behavior of prices, indicating, from the house buyers' standpoint,
a high level of substitutability, was observed in the following areas: 1. the Dan Metropolis
and the Central Region; 2. the Krayot and the northern region.
Additional findings of the study:
• The price elasticity of the "house size" characteristic is the most significant among
the characteristics examined in this study. The price elasticity relative to this characteristic
is also the highest.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 71
• Most physical characteristics of the houses, among them: house size, number of
rooms and house age, are characterized by a nonlinear relationship, signifying that the
marginal willingness to pay for an additional unit characteristic decreases.
• The majority of the environmental characteristics among them: the peripherality
index and the socioeconomic level (together with the floor characteristic) are characterized
by a linear relationship, that is, a relationship in which the willingness to pay for an
additional unit characteristic does not decrease.
• The characteristic with the greatest elasticity after house size is "number of rooms."
• The explained variance in Jerusalem is lower than in all other measured geographical
areas, due to the existence of other price influences of variables that are not observable in
the database.
Appendix A – Breakdown of House Area by Number of Rooms (1999-2009)
One-room apartment (8-65 sqm)
25
1.5-2 room apartment (10-77 sqm)
25
20.5
19.6
10
Maximum
‫ך מ ק סי מ לי‬value
‫ער‬
Minimum
‫ך מי ני מ לי‬value
‫ער‬
15
11.6
6.6
5.8
5
0.0
0.8
15
10
5
2.0 1.4
0.1 0.6
1.1 1.5 0.6
0.4
3.5-4 room apartment (55-142 sqm)
7.0
3.5
3.6 2.9 0.8 0.4 1.5
1.7 1.0
0.4 0.2
4 -0
1 4 -1 0
2 4 -2 0
3 4 -3 0
4 4 -4 0
5 4 -5 0
6 4 -6 0
7 4 -7 0
8 4 -8 0
9 4 -9 0
1 0 4 -1 0 0
1 1 4 -1 1 0
1 2 4 -1 2 0
1 3 4 -1 3 0
1 4 4 -1 4 0
150+
2.0 6.3
4.6
5 3.1
0.8 0.0 0.0 0.1 0.3 1.3
0.0 0.0 0.8 0.1 0.2 0.5
0
‫מ לי‬Minimum
‫ ע ר ך מי ני‬value
16
14
12
10
8
9.3
11.0
10.1
7.8
6.5
6
4.3
4 2.6
2
17.5
0.0
0.6
0.7
0.0 0.0
2.2
0.2 0.7
2.9
2.0
1.0 1.0
0.3
0.9
0
4 -0
9 -5
1 4 -1 0
1 9 -1 5
2 4 -2 0
2 9 -2 5
3 4 -3 0
3 9 -3 5
4 4 -4 0
4 9 -4 5
5 4 -5 0
5 9 -5 5
6 4 -6 0
6 9 -6 5
7 4 -7 0
7 9 -7 4
8 4 -8 0
8 9 -8 4
9 4 -9 0
9 9 -9 4
1 0 4 -1 0 0
1 0 9 -1 0 5
114+
10.2 8.7
Maximum
‫ך מ ק סי מ לי‬value
‫ער‬
‫לי‬Minimum
‫ ע ר ך מי ני מ‬value
11.9
18.4
18
Maximum
‫ע ר ך מ ק סי מ לי‬value
20
25.9
10
4 -0
9 -5
1 4 -1 0
1 9 -1 5
2 4 -2 0
2 9 -2 5
3 4 -3 0
3 9 -3 5
4 4 -4 0
4 9 -4 5
5 4 -5 0
5 9 -5 5
6 5 -6 0
+66
4 -0
9 -5
1 4 -1 0
1 9 -1 5
2 4 -2 0
2 9 -2 5
3 4 -3 0
3 9 -3 5
4 4 -4 0
4 9 -4 5
5 4 -5 0
5 9 -5 5
6 4 -6 0
6 9 -6 5
7 4 -7 0
7 9 -7 4
8 4 -8 0
+84
2.5-3 room apartment (37-100 sqm)
15
4.9
3.4
0
30
20
6.2
3.5
0.5 0.3 0.7
0
25
14.3
14.0
5.5 5.0
5.5
2.9
2.4
19.1
20
18.1
Minimum
‫ך מי ני מ לי‬value
‫ער‬
20
Maximum
‫ מ ק סי מ לי‬value
‫ערך‬
21.2
72
ISRAEL ECONOMIC REVIEW
4.5-5 room apartment (37-100 sqm)
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 73
Appendix B – Simulations for Sensitivity Analysis and Examination of Nonlinear
Relationships of the Explanatory Variables
Jerusalem
Intercept
***
Size
***
(1)
(2)
(3)
12.36763 ***
9.82099 ***
9.51253
(0.15762)
(0.26596)
(0.26879)
**
0.07406
(0.02969)
***
0.71395
(0.07856)
**
0.09313
(0.02869)
Ln Rooms
-0.00394
(0.00091)
Age
***
0.81969
(0.08137)
***
**
0.14865
(0.09237)
**
-0.00418
(0.00093)
0.77319 ***
0.79582
(0.08256)
(0.08351)
0.1574
(0.09386)
**
-0.00095
(0.01843)
**
0.01992
(0.00332)
**
0.01933
(0.00334)
0.14166
(0.09506)
-0.00427
(0.00093)
Ln Age
Socioeconomic Level
(4)
(5)
9.69894 ***
9.51072
(0.28426)
(0.29412)
0.01027
(0.0011)
Ln Size
Rooms
***
**
0.01894
(0.00336)
**
-0.00406
(0.01868)
0.01883
(0.00341)
**
Ln Socioeconomic Level
0.13647
(0.03463)
Potential Accessibility2
Ln Potential Accessibility
-0.01300
(0.00912)
-0.01436
(0.00736)
0.11078 **
(0.02616)
0.13394
(0.10764)
0.24304
(0.01777)
0.22188
(0.07546)
0.22002
(0.07621)
0.39782
(0.14885)
0.41119
(0.15109)
R Square
0.57677
0.56860
0.56755
0.55487
0.54206
Adjusted R Square
0.57145
0.56827
0.56199
0.54905
0.53668
Floor
-0.01471
(0.00756)
-0.01567
(0.00438)
-0.01616
(0.00766)
Ln Floor
Special Apartments
Status
**
**
0.15026
(0.10835)
**
0.16837
(0.10954)
**
0.1652
(0.11108)
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
74
ISRAEL ECONOMIC REVIEW
Appendix B – Simulations for Sensitivity Analysis and Examination of Nonlinear
Relationships of the Explanatory Variables
Jerusalem (cont.)
Intercept
***
(6)
9.51072
(0.29412)
***
(7)
9.48188
(0.2937)
***
(8)
(9)
(10)
9.75229 *** 9.71173 *** 9.51253
(0.27747)
(0.27644)
(0.28107)
Size
Ln Size
***
0.79582
(0.08351)
***
0.79613
(0.08367)
***
**
0.14166
(0.09506)
**
0.13925
(0.09515)
**
0.76514 *** 0.77554 *** 0.78576
(0.08333)
(0.08313)
(0.08326)
Rooms
Ln Rooms
0.14907
(0.09567)
**
0.14726
(0.09574)
**
0.13552
(0.09553)
Age
-0.00406
(0.01868)
Ln Age
-0.00440
(0.01872)
-0.00117
(0.00919)
-0.00186
(0.00918)
-0.00137
(0.00857)
Socioeconomic Level
Ln Socioeconomic Level
**
0.13647
(0.03463)
**
0.13708
(0.03468)
**
0.13382
(0.03484)
**
0.13336
(0.03489)
**
0.13268
(0.03498)
Potential Accessibility2
Ln Potential Accessibility
-0.01616
(0.00766)
Floor
-0.0185
(0.00943)
Ln Floor
Special Apartments
**
0.1652
(0.11108)
**
0.15557
(0.1121)
-0.01804
(0.00949)
**
-0.01907
(0.00942)
0.15577
(0.11279)
0.41119
(0.15109)
0.40773
(0.15136)
R Square
0.54206
0.54109
0.53347
0.53128
0.53060
Adjusted R Square
0.53668
0.53571
0.53145
0.52696
0.52599
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 75
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Tel Aviv
Intercept
Size
***
**
(1)
(2)
(3)
(4)
(5)
12.6287 ***
10.074 ***
9.93823 ***
10.0448 *** 9.53048
(0.0905)
(0.21826)
(0.22884)
(0.23111)
(0.25084)
0.00997
(0.0021)
Ln Size
Rooms
**
0.08111
(0.02742)
**
0.70521
(0.06569)
**
0.75169
(0.06919)
**
0.73676
(0.06896)
**
0.76008
(0.0708)
**
0.08002
(0.02552)
**
0.16465
(0.07554)
**
0.17611
(0.07463)
**
0.16973
(0.07647)
*
-0.00137
(0.00073)
*
-0.03576
(0.01215)
*
-0.03536
(0.06896)
0.03213 ***
0.03275 ***
(0.0033)
(0.00325)
0.03367
(0.00322)
***
0.34736
(0.0395)
Ln Rooms
Age
*
-0.0013
(0.00074)
*
-0.0012
(0.00074)
Ln Age
Socioeconomic Level
***
0.03259 ***
(0.00332)
Ln Socioeconomic Level
Potential Accessibility2
Ln Potential Accessibility
Floor
*
0.00948
(0.00646)
*
0.01286
(0.00449)
*
0.01332
(0.00445)
*
0.01279
(0.0044)
*
0.01306
(0.00451)
**
0.11078
(0.12616)
**
0.13394
(0.10764)
**
0.15026
(0.10835)
**
0.16837
(0.10954)
**
0.1652
(0.11108)
*
0.24304
(0.01777)
*
0.22188
(0.07546)
*
0.22002
(0.07621)
*
0.39782
(0.14885)
*
0.41119
(0.15109)
Ln Floor
Special Apartments
Status
R Square
0.71863
0.72842
0.72891
0.72792
0.71745
Adjusted R Square
0.71543
0.72793
0.72806
0.72637
0.71405
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
76
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Tel Aviv (cont.)
Intercept
***
(6)
9.53048
(0.25084)
***
(7)
(8)
9.51036 *** 9.75229 ***
(0.25309)
(0.27747)
(9)
(10)
9.71173 *** 9.51253
(0.27644)
(0.28107)
Size
Ln Size
**
0.76008
(0.0708)
**
0.77719
(0.07094)
**
0.78222
(0.07103)
**
0.81208
(0.07214)
**
0.80559
(0.07213)
**
0.16973
(0.07647)
**
0.1666
(0.07723)
**
0.17105
(0.07734)
**
0.16601
(0.07896)
**
0.16491
(0.07869)
* -0.03536
(0.06896)
*
-0.0407
(0.01241)
*
-0.0293
(0.00557)
*
-0.02915
(0.00568)
*
-0.0319
(0.00783)
Rooms
Ln Rooms
Age
Ln Age
Socioeconomic Level
Ln Socioeconomic Level
***
0.34736
(0.0395)
*
0.01306
(0.00451)
***
0.34849 *** 0.34471 ***
(0.03991)
(0.03977)
0.33874 *** 0.34093
(0.04053)
(0.04033)
*
0.00549
(0.00654)
*
0.00591
(0.00654)
0.00254
(0.00649)
**
0.23621
(0.06069)
Potential Accessibility2
Ln Potential Accessibility
Floor
Ln Floor
Special Apartments
Status
**
0.1652
(0.11108)
**
0.15557
(0.1121)
*
0.41119
(0.15109)
*
0.40773
(0.15136)
R Square
Adjusted R Square
0.71405
*
0.71295
0.71003
0.69997
0.70147
0.70951
0.70723
0.69662
0.69857
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 77
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Haifa
Intercept
***
Size
***
(1)
11.74196
(0.12396)
***
(2)
(3)
9.17140 *** 9.14743
(0.24559)
(0.24568)
0.11355
(0.02759)
***
0.73127 *** 0.74067
(0.07133)
(0.07193)
***
0.08978
(0.02626)
***
0.24534
(0.07824)
-0.0036 ***
(0.0009)
-0.00374
(0.0009)
Ln Rooms
Age
***
-0.00371
(0.00092)
***
Ln Age
Socioeconomic Level
***
***
(4)
(5)
9.25248 *** 8.54861
(0.25304)
(0.28181)
0.00886
(0.00104)
Ln Size
Rooms
***
0.04898
(0.00372)
***
0.04819 *** 0.04841
(0.00367)
(0.00361)
***
0.73083 *** 0.79205
(0.0721)
(0.07735)
***
0.26909 *** 0.25055
(0.07769)
(0.08413)
***
-0.07027 *** -0.04472
(0.01586)
(0.01709)
***
0.04941
(0.00358)
***
Ln Socioeconomic Level
0.45439
(0.04196)
Potential Accessibility2
Ln Potential Accessibility
-0.00359
(0.00587)
Floor
-0.00343
(0.00442)
-0.00369
(0.00442)
-0.00305
(0.00438)
-0.0045
(0.00473)
Ln Floor
Special Apartments
*
0.13728
(0.02173)
*
0.21544
(0.08259)
*
0.23951
(0.08239)
*
0.24299
(0.08218)
*
0.28807
(0.08931)
0.12557
(0.02574)
0.11058
(0.05085)
0.12865
(0.11093)
0.28667
(0.11847)
0.3288
(0.12754)
R Square
0.80469
0.81401
0.81737
0.81796
0.78739
Adjusted R Square
0.80092
0.81318
0.81319
0.81355
0.78318
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
78
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Haifa (cont.)
Intercept
***
(6)
(7)
(8)
(9)
(10)
8.54861 *** 8.54254 *** 8.32727 *** 8.24198 *** 8.32734
(0.28181)
(0.28083)
(0.27747)
(0.27229)
(0.2756)
Size
Ln Size
***
0.79205 *** 0.78992 *** 0.80845 *** 0.84245 *** 0.82647
(0.07735)
(0.07723)
(0.07755)
(0.07811)
(0.07801)
***
0.25055 *** 0.25323 *** 0.26349 *** 0.25730 *** 0.25198
(0.08413)
(0.08412)
(0.08468)
(0.08593)
(0.08536)
***
-0.04472 *** -0.04344 ***
-0.0346 *** -0.03418 *** -0.04888
(0.01709)
(0.01688)
(0.00645)
(0.00654)
(0.0104)
***
0.45439 *** 0.45253 *** 0.44975 *** 0.43210 *** 0.44660
(0.04196)
(0.04184)
(0.04212)
(0.04219)
(0.04224)
Rooms
Ln Rooms
Age
Ln Age
Socioeconomic Level
Ln Socioeconomic Level
Potential Accessibility2
Ln Potential Accessibility
-0.0045
(0.00473)
Floor
-0.0079
(0.00613)
Ln Floor
Special Apartments
*
0.28807
(0.08931)
*
0.28347
(0.08940)
-0.00714
(0.00617)
*
-0.00898
(0.00623)
0.29367
(0.08992)
0.3288
(0.12754)
0.32383
(0.12687)
R Square
0.78739
0.78811
0.78361
0.77525
0.77751
Adjusted R Square
0.78318
0.7839
0.77987
0.77218
0.77446
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 79
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Dan Metropolis
Intercept
***
Size
***
(1)
12.3841 ***
(0.08164)
***
**
(3)
8.6537 ***
(0.19441)
(4)
8.76501 ***
(0.19742)
(5)
8.15086
(0.20097)
0.40517 ***
(0.04632)
0.43813 ***
(0.04625)
0.43047 ***
(0.04611)
0.43651
(0.04651)
0.3919 **
(0.0488)
0.40163 **
(0.04816)
0.40208
(0.04861)
-0.04043 **
(0.00865)
-0.04091
(0.00872)
0.00582
(0.00074)
Ln Size
Rooms
(2)
8.7603 ***
(0.19714)
0.13802
(0.01837)
**
0.14839
(0.01674)
**
Ln Rooms
Age
**
-0.0018
(0.00066)
**
-0.0019
(0.0006)
**
-0.00212
(0.00059)
**
Ln Age
Socioeconomic Level
***
0.03794 ***
0.02998 ***
0.03021 ***
0.02972
(0.00238)
(0.00222)
(0.00219)
(0.00217)
***
Ln Socioeconomic Level
0.07894 *
(0.00076)
0.07054 *
(0.01427)
0.06085 *
(0.01409)
0.00071
(0.00583)
0.00881
(0.00255)
0.00921
(0.00251)
0.15254 **
(0.01341)
0.25003 **
(0.04476)
0.03528
(0.00746)
0.11837
(0.02677)
R Square
0.71262
Adjusted R Square
0.70903
Potential Accessibility
*
0.06883 *
(0.01401)
0.30428
(0.02379)
0.05007
(0.01388)
Ln Potential Accessibility
Floor
0.0089
(0.0025)
0.00904
(0.00252)
Ln Floor
Special Apartments
Status
**
0.27974 **
(0.0442)
0.27923 **
(0.04387)
0.28234
(0.04424)
0.13182
(0.02645)
0.10155
(0.06432)
0.1052
(0.06482)
0.77205
0.77078
0.77305
0.76736
0.77164
0.76779
0.76953
0.76534
***,**, *Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
80
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Dan Metropolis (cont.)
Intercept
***
(6)
(7)
(8)
(9)
(10)
5.33571 ***
5.35796 *** 5.27952 *** 5.14306 *** 5.21385
(0.32462)
(0.32751)
(0.32678)
(0.33156)
(0.33101)
Size
Ln Size
***
0.42125 ***
(0.04667)
0.45101 *** 0.45926 *** 0.48906 *** 0.47839
(0.04675)
(0.04682)
(0.04738)
(0.04728)
**
0.40103
(0.04857)
**
0.39943
(0.04912)
**
0.40403
(0.04923)
**
0.39868
(0.05008)
**
0.39246
(0.04972)
**
-0.04167
(0.00874)
**
-0.04622
(0.00871)
**
-0.0315
(0.00296)
**
-0.03092
(0.00301)
**
-0.03661
(0.00506)
0.30467 ***
0.31134 ***
(0.02365)
(0.02393)
0.31151 *** 0.31841 ***
(0.02402)
(0.02442)
0.31407
(0.0243)
2.2648
(0.13094)
2.23524
(0.13237)
Rooms
Ln Rooms
Age
Ln Age
Socioeconomic Level
Ln Socioeconomic Level
***
Potential Accessibility
Ln Potential Accessibility
*
*
*
*
2.24055
(0.13467)
*
2.24197
(0.13393)
0.00954
(0.00235)
Floor
0.0014
(0.00534)
Ln Floor
Special Apartments
2.24014
(0.13194)
**
0.28234
(0.04416)
**
0.27009
(0.0466)
-0.0007
(0.00536)
**
-0.0101
(0.00517)
0.26839
(0.04674)
0.10394
(0.06466)
0.12152
(0.065)
R Square
0.76713
0.76377
0.76159
0.75235
0.75455
Adjusted R Square
0.76508
0.76156
0.75954
0.75084
0.75302
Status
***,**, *Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 81
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Center
Intercept
***
Size
***
(1)
(2)
(3)
(4)
12.33373 ***
9.48304 ***
9.90622 ***
9.80521 ***
(0.05362)
(0.22041)
(0.14855)
(0.14547)
0.00336
(0.00055)
***
Ln Size
Rooms
**
0.15614 **
(0.01474)
0.55839 ***
0.48514 ***
0.51959 ***
0.55545
(0.06060)
(0.04275)
(0.04179)
(0.06291)
0.10191
(0.02078)
**
Ln Rooms
Age
**
-0.00539 **
(0.00058)
-0.00527 **
(0.00072)
0.34958
(0.04750)
**
0.03614 **
(0.00193)
0.03641 **
(0.00814)
**
0.34899
(0.04710)
**
0.35881
(0.07655)
**
-0.05892
(0.00565)
**
-0.05787
(0.01027)
**
0.36695
(0.03578)
-0.00583
(0.00054)
Ln Age
Socioeconomic Level
0.03595 **
(0.00177)
0.03545
(0.00176)
Ln Socioeconomic Level
Potential Accessibility
***
(5)
9.18631
(0.2235)
0.05209 ***
0.07142 ***
0.06418 ***
0.06229 ***
0.08645
(0.00019)
(0.01131)
(0.00376)
(0.00375)
(0.01129)
Ln Potential Accessibility
0.00778
(0.00335)
Floor
0.00659
(0.00480)
0.00341
(0.00224)
0.00411
(0.00222)
0.00456
(0.00482)
Ln Floor
Special Apartments
**
0.14822
(0.00748)
**
0.20166
(0.02262)
**
0.22841
(0.02353)
**
0.21831
(0.02334)
**
0.22099
(0.0234)
0.07844
(0.00491)
0.05227
(0.02692)
0.06876
(0.01630)
0.06714
(0.03854)
0.06158
(0.02667)
R Square
0.67891
0.73073
0.73916
0.73799
0.72905
Adjusted R Square
0.67738
0.72628
0.73777
0.73660
0.72455
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
82
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Center (cont.)
Intercept
***
(6)
8.98134
(0.15886)
***
(7)
(8)
(9)
8.92949 *** 8.41254 *** 8.48759
(0.15730)
(0.15579)
(0.15577)
***
(10)
8.64614
(0.15793)
Size
Ln Size
***
0.51779
(0.04246)
***
**
0.35122
(0.04796)
**
0.51730 *** 0.53962 *** 0.60201
(0.04198)
(0.04213)
(0.04247)
***
0.58205
(0.0426)
0.34250
(0.04887)
**
0.34530
(0.04873)
Rooms
Ln Rooms
0.35379
(0.04734)
**
0.32539
(0.04796)
**
Age
Ln Age
** -0.05874
(0.00572)
** -0.04125
(0.00564)
** -0.03913
(0.00214)
** -0.02873
(0.00218)
**
-0.02981
(0.00245)
**
**
**
**
0.38939
(0.01957)
**
0.38672
(0.01948)
0.38893 *** 0.41513 *** 0.41633 *** 0.43081
(0.02553)
(0.02814)
(0.02570)
(0.02607)
***
0.41362
(0.02561)
Socioeconomic Level
Ln Socioeconomic Level
0.39135
(0.01903)
0.38857
(0.01876)
0.39161
(0.01924)
Potential Accessibility
Ln Potential Accessibility
***
0.00454
(0.00226)
Floor
0.00726
(0.00318)
Ln Floor
Special Apartments
**
0.21704
(0.02376)
**
0.18625
(0.02484)
-0.01607
(0.00322)
**
-0.01561
(0.00304)
0.18329
(0.02521)
0.06993
(0.03898)
0.06375
(0.03858)
R Square
0.73432
0.73501
0.70243
0.71742
0.71914
Adjusted R Square
0.73289
0.73360
0.70120
0.71629
0.71804
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 83
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
South
Intercept
***
Size
***
(1)
(2)
(3)
(4)
12.0532 *** 9.61215 ***
9.50524 ***
9.59435 ***
(0.15762)
(0.26596)
(0.18113)
(0.18383)
0.00980
(0.00107)
***
Ln Size
Rooms
***
0.5711 ***
0.58698 ***
0.58213 ***
(0.0550)
(0.05654)
(0.05636)
***
**
-0.0053
(0.00064)
** -0.00470
(0.0005)
**
0.29750 ***
0.30370 ***
(0.06630)
(0.06606)
**
0.02372
(0.00321)
**
0.03361
(0.03098)
**
0.02986
(0.03076)
**
-0.05782
(0.0079)
**
-0.05771
(0.00794)
**
0.04354
(0.0306)
**
0.14955
(0.02399)
0.07257 ***
(0.00635)
0.07262
(0.00635)
0.00499
(0.00311)
0.00495
(0.00312)
Ln Socioeconomic Level
Potential Accessibility
***
0.30156
(0.06626)
-0.00464
(0.00057)
Ln Age
Socioeconomic Level
0.58812
(0.05645)
0.09386 *** 0.09400
(0.02087)
(0.01888)
Ln Rooms
Age
(5)
9.44285
(0.18686)
0.05082 *** 0.07092 *** 0.07175
(0.00508)
(0.00633)
(0.00634)
***
Ln Potential Accessibility
0.01267
(0.00366)
Floor
0.00497
(0.00316)
0.00491
(0.00313)
Ln Floor
Special Apartments
**
0.12209
(0.03769)
**
0.15003
(0.04476)
**
0.15026
(0.10835)
**
0.16837
(0.10954)
**
0.1658
(0.11108)
0.02489
(0.00582)
0.00102
(0.00046)
0.00109
(0.0004)
0.0009
(0.00045)
* -0.04415
(0.08342)
* -0.03303
(0.08286)
* -0.05052
(0.08305)
* -0.05750
(0.08175)
R Square
0.67888
0.73381
0.73260
0.73403
0.72909
Adjusted R Square
0.67578
0.73019
0.72897
0.73043
0.73271
Status
Confrontation Line Cities
0.00089
(0.00045)
*
-0.06560
(0.08209)
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
84
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
South (cont.)
Intercept
***
(6)
(7)
(8)
(9)
9.46591 *** 9.47700 *** 9.25310 ***
9.13085
(0.19092)
(0.19080)
(0.18540)
(0.18672)
***
(10)
9.26250
(0.19120)
Size
Ln Size
***
0.61390 *** 0.61701 *** 0.65178 ***
0.67797
(0.05742)
(0.05724)
(0.05737)
(0.05802)
***
0.64133
(0.05857)
***
0.28226 *** 0.28120 *** 0.26520 ***
0.28741
(0.06761)
(0.06754)
(0.06829)
(0.06927)
***
0.30106
(0.06935)
Rooms
Ln Rooms
Age
Ln Age
** -0.04803
(0.00820)
** -0.04875
(0.00813)
**
-0.03833
(0.00355)
**
-0.03709
(0.00359)
**
-0.03909
(0.00550)
**
**
**
0.15026
(0.02487)
**
0.15109
(0.0253)
**
0.16577
(0.02409)
0.16735 *** 0.16422 *** 0.15462 *** 0.15573
(0.01811)
(0.01830)
(0.01874)
(0.01854)
***
0.15717
(0.01824)
*
-0.06537
(0.04888)
Socioeconomic Level
Ln Socioeconomic Level
0.14415
(0.02453
0.14424
0.02459)
Potential Accessibility
Ln Potential Accessibility
***
0.00748
(0.00318)
Floor
0.00092
(0.00053)
Ln Floor
Special Apartments
**
0.1652
(0.11108)
**
0.65057
(0.1121)
0.00094
(0.0005)
**
0.00067
(0.00942)
0.16027
(0.11012)
0.00082
(0.00047)
0.00101
(0.00048)
* -0.08189
(0.08420)
* -0.08424
(0.08482)
* -0.07708
(0.08582)
R Square
0.71963
0.72126
0.71268
0.70079
0.69915
Adjusted R Square
0.71588
0.71750
0.70916
0.69750
0.69621
Status
Confrontation Line Cities
*
-0.07882
(0.08720)
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 85
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Sharon
Intercept
***
Size
***
(1)
12.09814
(0.07320)
0.00336
(0.00072)
***
Ln Size
Rooms
(2)
(3)
(4)
(5)
*** 10.26666 *** 9.37299 *** 9.37179 *** 8.94360
(0.25151)
0.20813)
(0.20504)
(0.20332)
***
0.14736
(0.02035)
***
0.49910 *** 0.51811 *** 0.52656 *** 0.54232
(0.0668)
(0.05902)
(0.05838)
(0.05877)
0.14046
(0.02169)
***
Ln Rooms
Age
**
-0.00463
(0.0009)
** -0.00313
(0.0011)
0.27207 *** 0.28234 *** 0.27239
(0.06760)
(0.06734)
(0.06788)
** -0.00362
(0.00081)
** -0.04620
(0.00895)
Ln Age
Socioeconomic Level
**
0.05041
(0.00274)
**
0.03945
(0.00450)
**
0.02683
(0.00332)
**
0.02216
(0.00331)
Ln Socioeconomic Level
Potential Accessibility
*
0.01313
(0.00061)
*
0.06221
(0.0122)
*
0.06623
(0.00991)
*
** -0.04575
(0.00902)
0.06603
(0.00988)
**
0.25589
(0.03579)
*
0.07744
(0.00950)
Ln Potential Accessibility
0.01433
(0.00427)
Floor
0.00211
(0.00386)
0.01214
(0.00316)
0.01205
(0.00315)
0.01175
(0.00318)
Ln Floor
Special Apartments
**
0.13843
(0.03967)
**
0.12739
(0.04120)
**
0.19943
(0.03016)
**
0.19238
(0.02994)
**
0.18970
(0.03020)
**
0.13477
(0.06392)
0.03349
(0.00730)
0.06763
(0.02912)
0.11645
(0.02430)
0.13808
(0.06336)
R Square
0.69581
0.75108
0.76724
0.76845
0.76436
Adjusted R Square
0.69315
0.75719
0.76493
0.76614
0.76202
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
86
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Sharon (cont.)
Intercept
***
(6)
(7)
(8)
(9)
(10)
7.84408 *** 7.87913 *** 7.75826 *** 7.51334 *** 7.67601
(0.21791)
(0.21908)
(0.21418)
(0.21429)
(0.21867)
Size
Ln Size
***
0.55181 *** 0.56396 *** 0.58144 *** 0.63438 *** 0.59781
(0.05791)
(0.05793)
(0.05791)
(0.05821)
(0.05874)
***
0.26166 *** 0.25642 *** 0.25717 *** 0.26175 *** 0.28024
(0.06689)
(0.06704)
(0.06741)
(0.06845)
(0.06835)
Rooms
Ln Rooms
Age
Ln Age
**
-0.03651
(0.00888)
**
-0.03874
(0.00887)
**
-0.02954
(0.00284)
** -0.02722
(0.00286)
**
-0.03384
(0.00493)
**
0.25613
(0.03472)
**
0.26050
(0.03476)
**
0.25674
(0.03489)
**
0.25804
(0.03553)
**
0.26170
(0.03553)
*
1.20027
(0.06099)
*
1.17062
(0.06188)
*
1.17035
(0.06223)
*
1.17821
(0.06314)
*
1.16674
(0.06253)
Socioeconomic Level
Ln Socioeconomic Level
Potential Accessibility
Ln Potential Accessibility
0.01086
(0.00313)
Floor
0.00439
(0.00381)
Ln Floor
Special Apartments
**
0.19582
(0.02975)
**
0.18027
(0.03148)
Status
**
0.14040
(0.06291)
**
0.14466
(0.06299)
0.00438
(0.00383)
**
0.00454
(0.00355)
0.17821
(0.03164)
R Square
0.76156
0.77046
0.76716
0.75562
0.75594
Adjusted R Square
0.75928
0.76819
0.76514
0.75384
0.75424
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 87
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
North
Intercept
(1)
(2)
(3)
*** 11.90868 *** 9.48304 ***
9.40410
(0.15762)
(0.22041)
(0.21605)
Size
***
***
**
0.12000
(0.02359)
**
0.55839 ***
0.55991
(0.06060)
(0.06313)
* -0.00388
(0.00078)
* -0.00337
(0.00072)
**
0.35005
(0.07644)
**
-0.00330
(0.00072)
**
0.55312
(0.06241)
***
0.55545
(0.06291)
**
0.35915
(0.07598)
**
0.35881
(0.07655)
0.03183
(0.00412)
**
0.04641
(0.00814)
**
0.05496
(0.05929)
**
**
0.02146
(0.00018)
**
0.08142
(0.01131)
**
0.08095
(0.01134)
**
** -0.05787
(0.01027)
0.05882
(0.00801)
Ln Socioeconomic Level
Potential Accessibility
(5)
9.18631
(0.22350)
***
** -0.05736
(0.01020)
Ln Age
Socioeconomic Level
***
0.10191
(0.02078)
Ln Rooms
Age
(4)
9.50345
(0.21643)
0.00551
(0.00082)
Ln Size
Rooms
***
0.08048
(0.01128)
**
0.26695
(0.03578)
**
0.08645
(0.01129)
Ln Potential Accessibility
0.00591
(0.00453)
Floor
-0.00059
(0.00480)
-0.00088
(0.00481)
-0.00122
(0.00479)
-0.00056
(0.00482)
Ln Floor
Special Apartments
**
0.03334
(0.00717)
Status
Confrontation Line Cities
0.13526
(0.10764)
***
**
0.13394
(0.10764)
0.15227
(0.02692)
**
0.15026
(0.10835)
**
0.16585
(0.02667)
0.20363 *** 0.20065 *** 0.19982
(0.02646)
(0.02921)
(0.02928)
0.16837
(0.10954)
**
0.16276
(0.02648)
***
0.19484
(0.02914)
0.1652
(0.11108)
0.16158
(0.02667)
***
0.20605
(0.02913)
R Square
0.68190
0.72073
0.72955
0.73301
0.72905
Adjusted R Square
0.67796
0.71628
0.72508
0.72859
0.72455
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
88
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
North (cont.)
Intercept
***
(6)
(7)
(8)
(9)
(10)
9.21592 *** 9.24566 *** 9.04883 *** 8.87820 *** 9.01614
(0.22151)
(0.22181)
(0.21153)
(0.21438)
(0.22057)
Size
Ln Size
***
0.55318 *** 0.54893 *** 0.57920 *** 0.61511 *** 0.58573
(0.06265)
(0.06284)
(0.06251)
(0.06364)
(0.06412)
Rooms
Ln Rooms
**
0.35888
(0.07628)
**
0.35906
(0.07645)
**
0.34698
(0.07694)
**
0.37942
(0.07857)
**
0.37842
(0.07877)
**
-0.05788
(0.01027)
** -0.04699
(0.01027)
** -0.03242
(0.00381)
** -0.03132
(0.00390)
** -0.04443
(0.00669)
**
0.27254
(0.03556)
**
0.26718
(0.03569)
**
0.27356
(0.03588)
**
0.26860
(0.03678)
**
0.29223
(0.03605)
**
0.22200
(0.02882)
**
0.22230
(0.02918)
**
0.22142
(0.02935)
**
0.22368
(0.03002)
**
0.20372
(0.02862)
Age
Ln Age
Socioeconomic Level
Ln Socioeconomic Level
Potential Accessibility
Ln Potential Accessibility
-0.00048
(0.00475)
Floor
-0.0185
(0.00943)
Ln Floor
Special Apartments
**
0.16347
(0.02655)
Status
Confrontation Line Cities
0.16382
(0.11108)
***
**
0.15836
(0.1121)
-0.00221
(0.00424)
**
-0.00817
(0.00419)
0.15569
(0.02699)
0.15773
(0.02679)
0.19391 *** 0.19394 *** 0.19027 *** 0.18644 *** 0.16115
(0.02841)
(0.02873)
(0.02889)
(0.02956)
(0.01837)
R Square
0.73006
0.73040
0.72507
0.70871
0.70685
Adjusted R Square
0.72565
0.72594
0.72097
0.70487
0.70334
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 89
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Krayot
Intercept
***
Size
***
(1)
(2)
11.88097 *** 9.48449 ***
(0.12482)
(0.30016)
***
**
0.11175
(0.03073)
**
0.66297 ***
(0.08434)
*
-0.00459
(0.00123)
0.63860 *** 0.63209
(0.08808)
(0.08663)
* -0.00365
(0.00120)
**
0.26786
(0.09850)
**
-0.00369
(0.00119)
**
0.29023
(0.09782)
** -0.04034
(0.01527)
Ln Age
Socioeconomic Level
**
(5)
9.20566
(0.31672)
0.03500
(0.00595)
**
***
0.63128
(0.08686)
**
0.29097
(0.09808)
0.07653
(0.02954)
Ln Rooms
Age
***
0.00692
(0.00114)
Ln Size
Rooms
(3)
(4)
9.53137 *** 9.61970
(0.30164)
(0.29803)
0.03370
(0.00572)
**
0.03394
(0.00569)
**
** -0.04945
(0.01530)
0.03360
(0.00564)
**
Ln Socioeconomic Level
0.33116
(0.05709)
Potential Accessibility2
Ln Potential Accessibility
-0.00320
(0.00688)
Floor
-0.00009
(0.00543)
-0.00033
(0.00540)
-0.00005
(0.00533)
-0.00011
(0.00534)
Ln Floor
Special Apartments
**
0.12491
(0.01759)
**
0.13514
(0.06128)
**
0.15459
(0.06065)
**
0.14785
(0.06007)
**
0.14636
(0.06021)
0.13372
(0.01203)
0.14241
(0.04237)
0.15024
(0.04226)
0.19249
(0.10811)
0.18689
(0.10832)
R Square
0.77631
0.79289
0.79446
0.79739
0.79624
Adjusted R Square
0.76977
0.78683
0.78845
0.79146
0.79028
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is considered
significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
90
ISRAEL ECONOMIC REVIEW
Simulations for Sensitivity Analysis and Examination of Nonlinear Relationships of
the Explanatory Variables
Krayot (cont.)
Intercept
***
(6)
(7)
(8)
9.20566 *** 9.21117 *** 9.01657
(0.31672)
(0.31585)
(0.30153)
***
(9)
(10)
8.92331 *** 9.04270
(0.30153)
(0.30802)
Size
Ln Size
***
0.63128 *** 0.63048 *** 0.66303
(0.08686)
(0.08649)
(0.08575)
***
0.69006 *** 0.66933
(0.08570)
(0.08611)
**
0.29097
(0.09808)
**
***
0.28135 *** 0.28202
(0.09951)
(0.09897)
**
-0.04945
(0.01530)
** -0.03101
(0.01515)
** -0.03528
(0.00515)
**
-0.03461
(0.05766)
** -0.04925
(0.00895)
**
0.33116
(0.05709)
**
0.33139
(0.05696)
**
0.32625
(0.05730)
**
0.31977
(0.05766)
**
-0.00725
(0.00660)
** -0.00628
(0.00662)
**
-0.00988
(0.00642)
0.13065
(0.06105)
**
Rooms
Ln Rooms
0.29300 *** 0.28317
(0.09793)
(0.09858)
Age
Ln Age
Socioeconomic Level
Ln Socioeconomic Level
0.32510
(0.05739)
Potential Accessibility2
Ln Potential Accessibility
-0.00011
(0.00534)
Floor
Ln Floor
Special Apartments
**
0.14636
(0.06021)
**
0.12933
(0.06144)
0.18689
(0.10832)
0.19819
(0.10744)
R Square
0.79624
0.79737
0.79290
0.78697
0.78810
Adjusted R Square
0.79028
0.79144
0.78774
0.78254
0.78388
Status
*,**,*** Represents the statistical significance at a level of 10%, 5% and 1%, respectively. A variable is
considered significant at a certain level if at least 80% of the cases were below the indicated level.
2
Potential accessibility is calculated at the level of the community and not at the level of the neighborhood, thus it
is not included in the simulations for geographical areas containing one city.
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 91
Appendix C – Examination of Stability and Time Trends of Regional Coefficients in
the Model (1999-2009)
Graph 15
Change in Coefficients over Time
Dan Metropolis
2007
2006
2005
2004
2006
2007
2008
2009
2006
2007
2008
2009
2005
2004
2005
2004
2003
SIZ
AGE
SAP
2002
2001
2000
Intercept
RMS
SEL
PAC
1999
2009
2008
2007
2006
2005
2004
2003
SIZ
AGE
SAP
Center
14
12
10
8
6
4
2
0
-2
SIZ
AGE
SAP
PAC
2002
2001
2000
1999
Intercept
RMS
SEL
CLC
2003
2002
2001
2000
1999
2009
2008
2007
2006
2005
2004
SIZ
AGE
SAP
Intercept
RMS
SEL
South
14
12
10
8
6
4
2
0
-2
SIZ
AGE
SAP
Haifa
14
12
10
8
6
4
2
0
-2
2003
2001
2000
1999
Intercept
RMS
SEL
PAC
2002
14
12
10
8
6
4
2
0
-2
2003
2002
2001
2000
Intercept
RMS
SEL
1999
2009
2008
2007
2006
2005
2004
2003
SIZ
AGE
SAP
2009
14
12
10
8
6
4
2
0
-2
2002
2001
2000
1999
Intercept
RMS
SEL
FLR
Jerusalem
2008
Tel Aviv
14
12
10
8
6
4
2
0
-2
92
ISRAEL ECONOMIC REVIEW
Appendix C (cont.) – Examination of Stability and Time Trends of Regional
Coefficients in the Model (1999-2009)
Graph 15
Change in Coefficients over Time
North
Krayot
SIZ
AGE
SAP
2009
2007
2006
2005
2004
2003
2002
2001
2000
1999
Intercept
RMS
SEL
2008
14
12
10
8
6
4
2
0
-2
2009
2008
2007
2006
2005
2004
SIZ
AGE
SAP
2003
2001
2000
Intercept
RMS
SEL
PAC
1999
2009
2008
2007
2006
2005
2004
SIZ
AGE
SAP
PAC
2002
14
12
10
8
6
4
2
0
-2
2003
2001
2000
1999
Intercept
RMS
SEL
CLC
2002
14
12
10
8
6
4
2
0
-2
Sharon
MEASURING LOCAL HOUSE PRICE MOVEMENTS IN ISRAEL AND ESTIMATING PRICE ELASTICITY 93
BIBLIOGRAPHY
Armknecht, Paul (1984). "Econometric issues of estimating hedonic functions", Journal of
Econometrics, 56, pp. 243-267.
Berndt E.R. & Rappaport N.J. (2001). "Price and Quality of Desktop and Mobile Personal
Computers: A Quarter-Century Historical Overview", American Economic Review,
91,2, pp. 268-273.
Brown, J.E., and D.E. Ethridge (1995). "Functional From Model Specification: An
Application to Hedonic Pricing", Agriculture and Resource Economy Review, 24, pp.
166-173.
Court, Andrew. (1939). "Hedonic Price Indexes with Automotive Examples", The
Dynamics of Automobile Demand Motors Corporation, New York, pp. 99-117.
Christensen, L.R. & Jorgenson D.W. and Lan L.J. (1975). "Transcendental logarithmic
utility functions", American Economic Review, 65, pp. 367-383.
Diewert, E. (2003). "Hedonic Regressions: A Review of Some Unresolved Issues",
presented at "International Conference in honor of Zvi Griliches: R&D, Education and
Productivity", Paris, August 25-27, 2003.
Diewert, E. (2007). "The Paris OECD-IMF Workshop on Real Estate Price Indexes:
Conclusions and Future Directions," Discussion Paper 07-01, Department of
Economics, University of British Columbia.
Diewert, E. & Heravi S. & Silver M. (2009). "Hedonic Imputation versus Time Dummy
Indexes" , in Price Index Concepts and Measurement, NBER, Studies in Income and
Wealth Vol. 70, University of Chicago Press, pp. 445-506.
Diewert, E. & Heravi S. & Silver M. (2009). "Durables and Owner Occupied Housing in a
Consumer Price Index" , in Price Index Concepts and Measurement, NBER, Studies in
Income and Wealth Vol. 70, University of Chicago Press, pp. 445-506.
Dullberger, Ellen R. (1989). "The Application of a Hedonic Model to the Quality-Adjusted
Price Index for Computer Processors", Technology and Capital Formation, Cambridge
pp.37-75.
Fisher, F & K.Shell (1972). "The Economic Theory of Price Indices: Two Essays on
Effects of Taste, Quality, and Technological Change. New York: Academic Press.
Freeman, A. M. (1979). "Hedonic Prices, Property Values and Measuring Environmental
Benefits", Scandinavian Journal of Economics, vol. 81, pp. 154-171.
Goodman, A. C. (1978). "Hedonic Prices, Price Indices and Housing Markets", Journal of
Urban Economics, vol. 34, pp. 57-79.
Griliches, Z (1961). "Hedonic Price Indexes for Automobiles: An Econometric Analysis of
Quality Change", The Price Statistics of the Federal Government, General Series no. 7,
Columbia University Press for NBER, New York, pp. 137-196.
Griliches, Z (1971). "Price Indexes and Quality Change, Studies in new method of
measurement", Cambridge: Harvard University Press.
Kain, John F. and John M. Quigley (1972). "Measuring the Value of Housing Quality",
Journal of the American Statistical Association, 62, pp. 263-77.
Lancaster, K. J. (1966). "A New Approach to Consumer Theory", Journal of Political
Economy, vol. 74, pp. 132-157.
94
ISRAEL ECONOMIC REVIEW
Melser, Daniel. (2005). "The Hedonic Regression Time-Dummy Method and Monotonicity
Axioms", Journal of Business & Economic Statistics, vol. 23, pp.485-492.
Okamoto, Masato & Tomohiko Sato. (2001). "Comparison of Hedonic Method and
Matched Models Method Using Scanner Data: The Case of PCs, TVs and Digital
Cameras", presented at the sixth Meeting of the International Working Group on Price
Indices, Canberra, Australia, April 2-6.
Pakes, A. (2003). "A Reconsideration of Hedonic Price Indices With an Application to
PC’s", American Economic Review, forthcoming.
Pfefferman, D. & Burrck, L. & Ben-Tuvia, S. (1989). "A Time Series Model For
Estimating Housing Price Indexes Adjusted For Changes In Quality” Analysis of Data
in Time. Ottawa: Statistics Canada.
Pfefferman, D. & Burrck, L. (1990). "Robust Small Area Combining Time Series and
Cross-Sectional Data", Survey Methodology Statistics Canada, vol.16, No. 2, pp. 217237.
Rosen, Sherwin. (1974). "Hedonic Prices and Implicit Markets: Product Differentiation in
Pure Competition" Journal of Political Economy 82(1), pp. 34-55.
Shepperd, S. (1999). "Hedonic analysis of housing markets, Handbook of Regional"
Journal of Urban and Economics, vol.3, Chapter 41, pp. 1595-1635.
Silver, M. and S. Heravi (2007). "The Difference Between Hedonic Imputation Indexes and
Time Dummy Hedonic Indexes", Journal of Business and Economic Statistics, 25(2),
pp. 239-246.
Triplett, Jack E. (1988). "Hedonic Functions and Hedonic Indexes" The New Palgraves
Dictionary of Economics, pp. 630-634.
Triplett, Jack E. (2004). "Handbook on Hedonic Indexes and Quality Adjustment in Price
Indexes" OECD Paper STI/ NA, Organization for Economic Co-operation and
Statistical Analysis of Science.
Waugh, F.V. (1928). "Quality Factors Influencing Vegetable Prices", Journal of Farm
Economics, vol. 10(2), pp.185-196.
Wallace, E.N. (1996). "Hedonic-Based Price Indexes for Housing: Theory, Estimation, and
Index Construction", FRBSF Economic Review vol.3, pp. 34-48.